Abstract
Background
The association between the atherogenic index of plasma (AIP), its modified indices (such as AIP-waist circumference [AIP-WC], AIP-waist-to-height ratio [AIP-WHtR], AIP-body mass index[AIP-BMI]), and incident stroke in individuals with cardiovascular-kidney-metabolic (CKM) stages 0–3 remains understudied. This study investigated these associations and their utility for risk stratification.
Methods
Data from 3697 China Health and Retirement Longitudinal Study (CHARLS) participants (≥ 45 years, CKM stages 0–3) were analyzed. Baseline, cumulative, and changes in AIP and its modified indices (AIP-WC, AIP-WHtR, AIP-BMI) were calculated. Logistic regression, Delong's test, integrated discrimination improvement (IDI), weighted quantile sum (WQS) regression, and mediation analysis were used to assess associations, predictive performance, component contributions, and mediation effects.
Results
Stroke occurred in 4.8% of participants. Under the fully adjusted Model 3: The third level of AIP, AIP-WHtR, AIP-WC, and AIP-BMI showed increased risks (ORs 1.58 [95% CI 1.05–2.38], 1.99 [95% CI 1.31–3.02], 1.99 [95% CI 1.31–3.02], and 1.92 [95% CI 1.26–2.92], respectively); The third level of cumulative AIP, AIP-WHtR, AIP-WC, and AIP-BMI showed elevated risks (ORs 1.79 [95% CI 1.19–2.69], 2.07 [95% CI 1.37–3.13], 2.01 [95% CI 1.33–3.04], and 1.92 [95% CI 1.27–2.89], respectively); The third category of AIP, AIP-WHtR, AIP-WC, and AIP-BMI changes showed risk increases of 2.28 (95% CI 1.46–3.55), 2.39 (95% CI 1.50–3.79), 2.56 (95% CI 1.61–4.07), and 2.22 (95% CI 1.38–3.56). Modified AIP indices (especially AIP-WHtR) demonstrated superior predictive ability than AIP alone. The association was amplified in advanced CKM (stages 2–3) but not significant in early CKM (stages 0–1). Triglycerides (TG) primarily drove the AIP-WHtR-stroke risk, which was partially mediated by estimated pulse wave velocity (ePWV) (6.48%).
Conclusions
AIP and its modified indices, especially AIP-WHtR, are significantly associated with incident stroke in CKM stages 0–3. Dynamically monitoring changes in these indices is crucial for stroke risk assessment and stratification, particularly in advanced CKM. TG primarily drives this risk, while ePWV partially mediates the AIP-WHtR-stroke link.
Graphic abstract
Supplementary Information
The online version contains supplementary material available at 10.1186/s12933-025-02784-8.
Keywords: Cardiovascular-kidney-metabolic syndrome, Stroke, Atherogenic index of plasma-waist circumference, Atherogenic index of plasma-waist-to-height ratio, Atherogenic index of plasma-body mass index, Estimated pulse wave velocity
Research insights
What is currently known about this topic?
Atherogenic index of plasma (AIP) increases stroke risk.
Higher incidence of stroke identified in cardiovascular-kidney-metabolic (CKM) population.
What is the key research question?
Whether AIP and its modified indices increase stroke risk in CKM stages 0–3 population remains unknown.
What is new?
AIP and its modified indices (especially AIP-waist-to-height ratio[AIP-WHtR]) are significantly associated with stroke risk in CKM stages 0-3, with AIP-WHtR showing superior performance (primarily driven by triglycerides and partially mediated by estimated pulse wave velocity).
How might this study influence clinical practice?
These indices aid in stroke risk stratification and personalized prevention.
Introduction
Cardiovascular-kidney-metabolic (CKM) syndrome is a systemic disorder caused by the pathophysiological interplay among metabolic risk factors, chronic kidney disease (CKD), and cardiovascular disease(CVD) [1, 2]. With the increasing prevalence of obesity and metabolic disorders worldwide, the incidence of CKM has been rising annually, posing a significant challenge to global public health [3]. The clinical burden of CVD varies disproportionately across different stages of CKM syndrome, underscoring the necessity of considering metabolism, kidney function, and cardiovascular systems as a unified entity [4–6]. The American Heart Association (AHA) emphasizes that research in populations with CKM syndrome stages 0–3 should focus more on the prevention of CVD onset [2].
Stroke, a common and severe complication in CKM patients, is closely associated with multiple metabolic and inflammatory biomarkers [7, 8]. In recent years, the atherogenic index of plasma (AIP), a novel lipid metabolism biomarker, has gained significant attention due to its strong correlation with CVD and metabolic syndrome [9, 10]. AIP reflects the risk of atherosclerosis by calculating the logarithmic ratio of triglycerides (TG) to high-density lipoprotein cholesterol (HDL) [11]. Previous studies have shown that combining obesity-related indicators such as body mass index (BMI), waist circumference (WC), and waist-to-height ratio (WHtR) can provide a more comprehensive assessment of cardiovascular and metabolic risks. However, these studies have primarily focused on the combination of triglyceride-glucose (TyG) with obesity-related indicators [12–17]. Nevertheless, research on the relationship between AIP and its modified indices—AIP-WC, AIP-WHtR, and AIP-BMI—and stroke risk in CKM stage 0–3 populations remains insufficient.
Therefore, this study aims to investigate the association between AIP, AIP-WC, AIP-WHtR, and AIP-BMI and incident stroke in CKM stage 0–3 populations. By analyzing the distribution characteristics of these indices across different CKM stages and their association with stroke, we aim to perform risk stratification. This will provide a scientific basis for the early identification of high-risk populations for stroke and the development of personalized intervention strategies.
Methods
Study population
This is a cohort study based on the China Health and Retirement Longitudinal Study (CHARLS), which recruited 17,708 participants from 28 provinces in China (2011–2012 baseline survey, Wave 1, 80.5% response rate) [18]. Follow-up surveys were conducted in 2013 (Wave 2), 2015 (Wave 3), 2018 (Wave 4), and 2020 (Wave 5), with blood samples collected during Wave 1 and Wave 3. Participants aged 45 and older with complete data on HDL, TG, WC, WHtR, and BMI were included in this analysis. Individuals with a history of stroke or heart disease prior to 2015 were excluded (Fig. 1). The study was approved by the Peking University IRB (IRB00001052-11015) and adhered to the STROBE guidelines. All participants provided written informed consent [19].
Fig. 1.
Flowchart of the study population. Blood samples were collected at wave 1 for baseline and at wave 3 for follow-up, while wave 2 did not involve blood sample collection and was excluded from the analysis. TG triglyceride; HDL high-density lipoprotein cholesterol; BMI body mass index; WC waist circumference; WHtR waist-to-height ratio
Measures
Participants were categorized into tertiles based on AIP, AIP-WC, AIP-WHtR, and AIP-BMI values, calculated using baseline (Wave 1) and cumulative data (Waves 1–3). The indices were calculated as follows: (1) AIP = Log [TG (mg/dl) /HDL(mg/dl)] [20]; (2) BMI = body mass(kg)/height2 (m2); (3)WHtR = WC/height; (4) AIP-WC = AIP × WC; AIP-WHtR = AIP × WHtR; AIP-BMI = AIP × BMI. We calculated cumulative AIP-WC, AIP-WHtR, and AIP-BMI based on a formula derived from cumulative AIP [21]: (AIP2012 + AIP2015)/2 × time (2015–2012).
CVD was defined as a self-reported history of heart disease and/or stroke [22–24]. Regarding the diagnosis of heart disease, the participants were verbally assessed and asked, “Has your physician ever informed you that you had a heart attack, angina, coronary artery disease, heart failure, or other heart-related problems? Furthermore, to ascertain whether a stroke had been diagnosed, the inquiry, “Has your doctor ever told you that you have been diagnosed with a stroke?” was posed.
The primary outcome was incident stroke, and the date of stroke diagnosis was recorded as the period between the most recent prior interview and the interview where the stroke was reported [25, 26].
Covariates were selected based on prior studies and clinical expertise [25, 26]. Categorical variables included sex, residence, marital status, education, smoking, drinking status, and diabetes. Continuous variables included age, systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol (TC), low-density lipoprotein cholesterol (LDL), and estimated glomerular filtration rate (eGFR). eGFR was calculated using the Chinese-adapted Modification of Diet in Renal Disease (MDRD) equation based on serum creatinine (SCr): eGFR = 175*SCr^(− 1.234) * Age^(− 0.179)*0.79[if female] [27]. Diabetes was defined as follows: fasting blood glucose (FBG) ≥ 125 mg/dL or Glycated Hemoglobin (HbA1c) ≥ 6.5%, self-reported diagnosis of diabetes, or the use of medications for diabetes.
Definition of CKM syndrome stages 0 to 3
The CKM syndrome stages are defined as follows, as outlined in the AHA Presidential Advisory Statement [2]: Stage 0: The absence of CKM syndrome risk factors. Stage 1: Abdominal obesity and/or prediabetes. Stage 2: Metabolic conditions (type 2 diabetes mellitus [T2DM], hypertension, hypertriglyceridemia) or kidney disease. Stage 3: Early cardiovascular disease in individuals with obesity, metabolic conditions, or kidney disease.The Framingham risk score estimates CVD risk based on various risk factors (see Table S1 and S2 for details).
For this study, participants with clinical CVD (CKM stage 4) were excluded at Waves 1 and 3. Specific staging criteria for CKM stages 0–3 are detailed in Table S3 and diagnostic procedures for overweight, abdominal obesity, hypertension, diabetes, and metabolic syndrome are outlined in Table S4.
Statistical analysis
Statistical analysis and modeling were performed in 2024. Continuous variables were presented as medians (interquartile ranges [IQRs]) or means (standard deviations [SD]). Data distribution was assessed using normality tests. The Kruskal–Wallis rank sum test was applied for non-normally distributed variables, while analysis of variance (ANOVA) was used for normally distributed variables. Categorical variables were described as frequencies and percentages. Fisher’s exact test was applied for categorical variables with expected frequencies < 5, and the Chi-squared test was used for expected frequencies ≥ 5 [28].
K-means clustering with Euclidean distance was used to group patients based on AIP, AIP-WC, AIP-WHtR, and AIP-BMI values from 2012 and 2015. K-means was chosen for its computational efficiency and ability to produce clear visualizations [29]. This centroid-based clustering method minimizes the sum of squared distances within each cluster [29, 30]. The elbow method determined the optimal number of clusters as 3 (Fig.S1). Class 1 showed low levels with a gradual increase; Class 2 maintained moderate levels with some upward trend; Class 3 displayed high levels with a slight decline. These trends were statistically significant over the period from 2012 to 2015 (P < 0.003). Detailed category analysis data and time trends can be found in Fig. 2A, Fig. 2B, and Table S5. Figure 2C and Fig. 2D show the distribution of AIP, AIP-WC, AIP-WHtR, and AIP-BMI across different classes, each following a normal distribution with significant differences in mean values.
Fig. 2.
Clustering of changes in AIP, AIP-WC, AIP-WHtR, and AIP-BMI from 2012 to 2015. Three clusters (Class 1, Class 2, and Class 3) were identified using the K-means method with Euclidean distance. Each cluster is represented by a unique color and shape across all figures to distinguish between groups and highlight variations over time. A Scatter plots comparing standardized values of AIP, AIP-WC, AIP-WHtR, and AIP-BMI between 2012 (X-axis) and 2015 (Y-axis). Standardization adjusts the data to a mean of 0 and a standard deviation of 1, facilitating comparison across years. B Line plots showing the mean values of AIP, AIP-WC, AIP-WHtR, and AIP-BMI for the three clusters from 2012 to 2015. Each indicator is presented at two time points (2012 and 2015), illustrating temporal trends and highlighting differences in mean values across clusters. C, D Density plots illustrating the distributions of AIP, AIP-WC, AIP-WHtR, and AIP-BMI within the three clusters in 2012 and 2015. These plots reveal significant differences in mean levels and variability among clusters, as well as shifts in distributions of indicators over time, allowing for visual comparison across the 2 years. AIP atherogenic index of plasma; WC waist circumference; WHtR waist-to-height ratio; BMI body mass index
Three models were used to assess the impact of covariates, with the first two applied for preliminary analyses to demonstrate associations with and without partial adjustments, ensuring the robustness of the final model (Model 3). Model 1 was unadjusted. Model 2 was adjusted for age, sex, marital status, residence, educational level, drinking status, and smoking status. Model 3 included the variables in Model 2, with additional adjustments for diabetes, SBP, DBP, TC, LDL, and eGFR.
The association between AIP and its modified indices with incident stroke was examined using binary logistic regression models to calculate odds ratios (ORs) with 95% confidence intervals (CIs). The False Discovery Rate (FDR) correction using the Benjamini–Hochberg method was applied in Model 3 to control for type I errors due to multiple comparisons. To investigate linear trends, the median value of each tertile was modeled to test ordered relationships across tertiles of AIP and its modified indices. Potential nonlinear associations were also explored using a restricted cubic spline (RCS) regression model with 4 knots at the 5th, 35th, 65th, and 95th percentiles of AIP and its modified indices values and their cumulative measures. The 5th percentile served as the reference.
Subgroup analyses were used to investigate whether demographic and clinical characteristics (age, sex, marital status, residence, educational level, smoking status, drinking status, CKM stages, and history of hypertension, heart disease, dyslipidemia, diabetes, and kidney disease) moderated the association between AIP and its modified indices with incident stroke. Imbalanced CKM stage 0–3 samples (Table 1) prompted reclassification into: early CKM (0–1: no/only metabolic risks); advanced CKM (2–3: metabolic disorders/subclinical CVD), enhancing effect size comparability through clinical alignment. Interaction terms and likelihood ratio tests were used to evaluate P values for interaction. P values for trends were calculated by modeling the median values of each quantile to test ordered relations across quantiles of AIP and its modified indices.
Table 1.
Baseline Characteristics of 3697 participants according to AIP at Wave1 (2011–2012)
| Variables | Total (n = 3697) | AIP2012 | P value | ||
|---|---|---|---|---|---|
| Tertile 1 [− 0.501, 0.176) | Tertile 2 [0.176,0.46) | Tertile 3 [0.46,2.122] | |||
| (n = 1232) | (n = 1232) | (n = 1233) | |||
| Age, Mean ± SD | 58.5 ± 8.6 | 58.9 ± 8.8 | 58.7 ± 8.6 | 58.0 ± 8.5 | 0.016 |
| Sex, n (%) | < 0.001 | ||||
| Female | 1980 (53.6) | 603 (48.9) | 676 (54.9) | 701 (56.9) | |
| Male | 1717 (46.4) | 629 (51.1) | 556 (45.1) | 532 (43.1) | |
| Marital Status, n (%) | 0.212 | ||||
| Married | 3186 (86.2) | 1047 (85) | 1061 (86.1) | 1078 (87.4) | |
| Other | 511 (13.8) | 185 (15) | 171 (13.9) | 155 (12.6) | |
| Residence, n (%) | < 0.001 | ||||
| Rural | 2492 (67.4) | 894 (72.6) | 832 (67.5) | 766 (62.1) | |
| Urban | 1205 (32.6) | 338 (27.4) | 400 (32.5) | 467 (37.9) | |
| Educational level, n (%) | 0.833 | ||||
| College or above | 42 (1.1) | 13 (1.1) | 16 (1.3) | 13 (1.1) | |
| Middle or high school | 1051 (28.4) | 347 (28.2) | 344 (27.9) | 360 (29.2) | |
| Primary school | 1550 (41.9) | 528 (42.9) | 503 (40.8) | 519 (42.1) | |
| No formal education | 1054 (28.5) | 344 (27.9) | 369 (30) | 341 (27.7) | |
| Smoking status, n (%) | 0.013 | ||||
| Current | 1140 (30.8) | 423 (34.4) | 363 (29.5) | 354 (28.7) | |
| Ever | 272 (7.4) | 93 (7.6) | 83 (6.7) | 96 (7.8) | |
| Never | 2284 (61.8) | 715 (58.1) | 786 (63.8) | 783 (63.5) | |
| Drinking status, n (%) | < 0.001 | ||||
| Current | 1197 (32.4) | 478 (38.8) | 346 (28.1) | 373 (30.3) | |
| Ever | 275 (7.4) | 77 (6.2) | 103 (8.4) | 95 (7.7) | |
| Never | 2225 (60.2) | 677 (55) | 783 (63.6) | 765 (62) | |
| Hypertension, n (%) | 1986 (53.8) | 548 (44.6) | 663 (53.9) | 775 (62.9) | < 0.001 |
| Dyslipidemia, n (%) | 278 (7.7) | 55 (4.5) | 84 (7) | 139 (11.5) | < 0.001 |
| Diabetes, n (%) | 644 (17.4) | 134 (10.9) | 185 (15) | 325 (26.4) | < 0.001 |
| Kidney Disease, n (%) | 204 (5.5) | 67 (5.5) | 67 (5.5) | 70 (5.7) | 0.958 |
| Stroke, n (%) | 176 (4.8) | 42 (3.4) | 61 (5) | 73 (5.9) | 0.013 |
| CKM Stages, n (%) | < 0.001 | ||||
| Stage 0 | 307 (8.3) | 215 (17.5) | 89 (7.2) | 3 (0.2) | |
| Stage 1 | 1105 (29.9) | 417 (33.8) | 522 (42.4) | 166 (13.5) | |
| Stage 2 | 2102 (56.9) | 578 (46.9) | 583 (47.3) | 941 (76.3) | |
| Stage 3 | 183 (4.9) | 22 (1.8) | 38 (3.1) | 123 (10) | |
| SBP2012 (mmHg), Mean ± SD | 128.8 ± 20.3 | 125.6 ± 19.5 | 129.0 ± 20.7 | 131.9 ± 20.1 | < 0.001 |
| DBP2012 (mmHg), Mean ± SD | 75.3 ± 11.8 | 73.4 ± 11.5 | 75.4 ± 11.9 | 77.3 ± 11.8 | < 0.001 |
| WHtR2012, Mean ± SD | 0.5 ± 0.1 | 0.5 ± 0.1 | 0.5 ± 0.1 | 0.6 ± 0.1 | < 0.001 |
| WHtR2015, Mean ± SD | 0.5 ± 0.1 | 0.5 ± 0.1 | 0.5 ± 0.1 | 0.6 ± 0.1 | < 0.001 |
| WC2012(cm), Mean ± SD | 84.2 ± 12.1 | 80.2 ± 10.4 | 84.4 ± 11.7 | 88.1 ± 12.9 | < 0.001 |
| WC2015(cm), Mean ± SD | 84.5 ± 14.0 | 79.2 ± 14.9 | 84.9 ± 12.9 | 89.5 ± 11.9 | < 0.001 |
| BMI2012(kg/m2), Mean ± SD | 23.5 ± 3.8 | 22.1 ± 3.4 | 23.5 ± 3.8 | 25.0 ± 3.7 | < 0.001 |
| BMI2015(kg/m2), Mean ± SD | 23.8 ± 3.9 | 22.4 ± 3.7 | 23.8 ± 3.8 | 25.1 ± 3.9 | < 0.001 |
| TG2012(mg/dl), Median (IQR) | 102.7 (72.6, 149.6) | 64.6 (54.0, 75.2) | 102.2 (88.5, 118.6) | 182.3 (146.0, 242.5) | < 0.001 |
| TG2015(mg/dl), Mean ± SD | 137.1 ± 86.5 | 92.7 ± 41.8 | 126.1 ± 63.6 | 192.4 ± 107.3 | < 0.001 |
| HDL2012(mg/dl), Mean ± SD | 51.5 ± 15.3 | 64.9 ± 14.2 | 50.7 ± 9.5 | 39.1 ± 8.8 | < 0.001 |
| HDL2015(mg/dl), Mean ± SD | 52.1 ± 12.0 | 58.7 ± 13.3 | 51.4 ± 9.9 | 46.1 ± 8.8 | < 0.001 |
| TC2012(mg/dl), Mean ± SD | 194.4 ± 38.8 | 189.5 ± 34.1 | 192.7 ± 37.5 | 201.2 ± 43.3 | < 0.001 |
| LDL2012(mg/dl), Mean ± SD | 117.0 ± 34.7 | 113.6 ± 29.8 | 122.6 ± 33.9 | 114.8 ± 39.2 | < 0.001 |
| HBA1C2012, Mean ± SD | 5.3 ± 0.8 | 5.2 ± 0.7 | 5.2 ± 0.7 | 5.4 ± 1.0 | < 0.001 |
| eGFR (ml/ min/1.73 m2), Mean ± SD | 111.0 ± 29.8 | 112.5 ± 26.5 | 110.3 ± 27.3 | 110.1 ± 35.0 | 0.076 |
| AIP2012, Median (IQR) | 0.3 (0.1, 0.5) | 0.0 (− 0.1, 0.1) | 0.3 (0.2, 0.4) | 0.6 (0.5, 0.8) | < 0.001 |
| AIP2015, Median (IQR) | 0.3 (0.2, 0.5) | 0.2 (0.0, 0.3) | 0.3 (0.2, 0.5) | 0.6 (0.4, 0.7) | < 0.001 |
| AIP-WC2012, Median (IQR) | 25.7 (8.1, 46.9) | 1.9 (− 5.9, 8.2) | 26.2 (19.9, 32.7) | 57.1 (46.9, 74.1) | < 0.001 |
| AIP-WC2015, Median (IQR) | 29.0 (12.7, 48.4) | 12.4 (2.7, 24.4) | 28.9 (16.8, 43.3) | 49.8 (34.7, 66.7) | < 0.001 |
| AIP-WHtR2012, Median (IQR) | 0.2 (0.1, 0.3) | 0.0 (0.0, 0.1) | 0.2 (0.1, 0.2) | 0.4 (0.3, 0.5) | < 0.001 |
| AIP-WHtR2015, Median (IQR) | 0.2 (0.1, 0.3) | 0.1 (0.0, 0.2) | 0.2 (0.1, 0.3) | 0.3 (0.2, 0.4) | < 0.001 |
| AIP-BMI2012, Median (IQR) | 7.2 (2.3, 13.2) | 0.5 (− 1.6, 2.3) | 7.2 (5.4, 9.0) | 16.0 (13.2, 20.8) | < 0.001 |
| AIP-BMI2015, Median (IQR) | 8.1 (3.7, 13.5) | 3.5 (0.8, 6.7) | 8.0 (4.7, 12.1) | 13.8 (9.6, 19.0) | < 0.001 |
Data following a normal distribution, expressed as mean ± standard deviation, were analyzed using a One-way ANOVA. Proportions were compared using Pearson's Chi-squared test, and non-normally distributed data were analyzed using the Kruskal–Wallis rank sum test
AIP atherogenic index of plasma; SBP systolic blood pressure; DBP diastolic blood pressure; TC total cholesterol; HDL high-density lipoprotein cholesterol; LDL low-density lipoprotein cholesterol; eGFR estimated glomerular filtration ratio; TG triglyceride; WHtR waist-to-height ratio; WC waist circumference; BMI body mass index. CKM cardio-kidney-metabolic syndrome
To evaluate the predictive performance of AIP and its modified indices for incident stroke, a receiver operating characteristic (ROC) curve analysis was conducted, and the area under the curve (AUC) was calculated. Pairwise comparisons using Delong's test were used to identify indices with superior predictive performance. Integrated discrimination improvement (IDI) index were used to further evaluate the incremental predictive value.
The AIP and its modified indices were calculated using a formula that incorporates TG, HDL, or combined with an obesity indicator (e.g., WC, WHtR, or BMI). To quantify the relative contributions of TG, HDL, and the selected obesity indicator to stroke prediction, we utilized a WQS regression model, employing bootstrap resampling methods for 1000 iterations. This approach assigns weights to each variable, constrained between 0 and 1 [31], reflecting their importance in predicting incident stroke. Higher weights indicate greater significance of a variable in the overall effect.
Previous studies demonstrated that arterial stiffness was recognized as an important contributing factor for cardiovascular diseases and adverse cardiovascular events (including stroke) [32, 33]. Therefore, arterial stiffness (represented by ePWV [34]) was selected as a potential mediating factor for the associations of AIP and its modified indices and incident stroke. ePWV was calculated from age and mean blood pressure (MBP) [34, 35]:9.587 − 0.402 × age + 4.560 × 10−3 × age2 − 2.621 × 10−5 × age2 × MBP + 3.176 × 10−3 × age × MBP − 1.832 × 10−2 × MBP. MBP was calculated as DBP + 0.4 × (SBP– DBP).
In addition, sensitivity analyses were conducted on the data prior to multiple imputation to verify the robustness of the results.
All statistical analyses were conducted using R (v4.4.1), Free Statistics software (version 2.1.1; Beijing Free Clinical Medical Technology Co., Ltd.).
Results
The study included 3,697 participants (mean age: 58.5 years, 53.6% female, 86.2% married, 67.4% rural residents). Educational attainment was divided between primary school (41.9%) and middle/high school education (28.4%). Lifestyle factors included 30.8% current smokers and 32.4% current drinkers. Health conditions were prevalent, with hypertension (53.8%), dyslipidemia (7.7%), diabetes (17.4%), chronic kidney disease (5.5%).CKM stages were primarily distributed as follows: Stage 2 (56.9%), Stage 1 (29.9%), Stage 3 (4.9%), and Stage 0 (8.3%). Mean blood pressure readings were 128.8 mmHg (SBP) and 75.3 mmHg (DBP). Anthropometric metrics included WHtR (0.5), WC (84.2 cm), and BMI (23.5 kg/m2). Key lipid and metabolic metrics were: TG (median: 102.7 mg/dl), HDL (51.5 mg/dl), TC (194.4 mg/dl), LDL (117.0 mg/dl), HbA1c (5.3%), and eGFR (111.0 ml/min/1.73 m2, P = 0.076). Significant differences (P < 0.001) were observed across AIP tertiles for SBP, DBP, TC, HDL-C, LDL-C, HbA1c, FBG, TG, WC, WHtR, and BMI (Table 1).
Among 3,697 participants, 176 (4.8%) experienced stroke between 2015 and 2020. Table 2 reveals significant associations between AIP and its modified indices (baseline, cumulative, and changes) and incident stroke. Under the fully adjusted Model 3: AIP: The third level showed a 1.58-fold increased stroke risk (95% CI 1.05–2.38). AIP-WHtR: The third level significantly increased stroke risk to 1.99-fold (95% CI 1.31–3.02). AIP-WC: The third level showed a 1.99-fold risk increase (95% CI 1.31–3.02). AIP-BMI: The third level increased risk to 1.92-fold (95% CI 1.26–2.92). The third level of cumulative AIP, AIP-WHtR, AIP-WC, and AIP-BMI showed elevated risks (ORs 1.79 [95% CI 1.19–2.69], 2.07 [95% CI 1.37–3.13], 2.01 [95% CI 1.33–3.04], and 1.92 [95% CI 1.27–2.89], respectively).The third category of AIP, AIP-WHtR, AIP-WC, and AIP-BMI changes showed stroke risk increases of 2.28 (95% CI 1.46–3.55), 2.39 (95% CI 1.50–3.79), 2.56 (95% CI 1.61–4.07), and 2.22 (95% CI 1.38–3.56), respectively. After applying the FDR, the above results remained statistically significant (P < 0.05) (Table 2). Similarly, the results were largely consistent after multiple imputation (Table S6 and Table S7).
Table 2.
Associations of different levels of AIP and its modified indices with incident stroke
| Variable | Total (n.) | Event (%) | Model 1 | Model 2 | Model 3 | P value FDR | |||
|---|---|---|---|---|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | ||||
| AIP2012 | |||||||||
| AIP2012 | 3697 | 176 (4.8) | 2.33 (1.55–3.51) | < 0.001 | 2.59 (1.71–3.94) | < 0.001 | 2.69 (1.57–4.61) | < 0.001 | 0.001 |
| Tertile 1 [− 0.501, 0.176) | 1232 | 42 (3.4) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Tertile 2 [0.176, 0.460) | 1232 | 61 (5) | 1.48 (0.99–2.20) | 0.057 | 1.51 (1.01–2.26) | 0.047 | 1.40 (0.93–2.11) | 0.111 | 0.133 |
| Tertile 3 [0.460, 2.122] | 1233 | 73 (5.9) | 1.78 (1.21–2.63) | 0.003 | 1.89 (1.27–2.8) | 0.002 | 1.58 (1.05–2.38) | 0.029 | 0.046 |
| P for Trend | 0.004 | 0.002 | 0.030 | ||||||
| AIP-WHtR2012 | |||||||||
| AIP-WHtR2012 | 3697 | 176 (4.8) | 4.99 (2.51–9.89) | < 0.001 | 5.94 (2.94–12) | < 0.001 | 6.55 (2.61–16.41) | < 0.001 | 0.001 |
| Tertile 1 [− 0.304, 0.090) | 1232 | 37 (3) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Tertile 2 [0.090, 0.248) | 1232 | 59 (4.8) | 1.62 (1.07–2.47) | 0.023 | 1.67 (1.09–2.55) | 0.017 | 1.57 (1.02–2.41) | 0.038 | 0.052 |
| Tertile 3 [0.248, 1.227] | 1233 | 80 (6.5) | 2.24 (1.51–3.34) | < 0.001 | 2.37 (1.58–3.56) | < 0.001 | 1.99 (1.31–3.02) | 0.001 | 0.003 |
| P for Trend | < 0.001 | < 0.001 | 0.001 | ||||||
| AIP-WC2012 | |||||||||
| AIP-WC2012 | 3697 | 176 (4.8) | 1.01 (1.01–1.01) | < 0.001 | 1.01 (1.01–1.02) | < 0.001 | 1.01 (1.01–1.02) | < 0.001 | 0.001 |
| Tertile 1 [− 43.157, 14.136] | 1232 | 37 (3) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Tertile 2 [14.238, 38.842] | 1232 | 60 (4.9) | 1.65 (1.09–2.51) | 0.018 | 1.69 (1.11–2.58) | 0.014 | 1.58 (1.03–2.42) | 0.037 | 0.052 |
| Tertile 3 [38.885, 212.217] | 1233 | 79 (6.4) | 2.21 (1.48–3.29) | < 0.001 | 2.35 (1.57–3.53) | < 0.001 | 1.99 (1.31–3.02) | 0.001 | 0.003 |
| P for Trend | < 0.001 | < 0.001 | 0.001 | ||||||
| AIP-BMI2012 | |||||||||
| AIP-BMI2012 | 3697 | 176 (4.8) | 1.03 (1.02–1.05) | < 0.001 | 1.04 (1.02–1.06) | < 0.001 | 1.04 (1.02–1.06) | < 0.001 | 0.001 |
| Tertile 1 [− 11.826, 3.988] | 1232 | 38 (3.1) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Tertile 2 [4.002, 11.009] | 1232 | 61 (5) | 1.64 (1.08–2.47) | 0.019 | 1.68 (1.11–2.55) | 0.015 | 1.56 (1.02–2.38) | 0.039 | 0.052 |
| Tertile 3 [11.014, 61.157] | 1233 | 77 (6.2) | 2.09 (1.41–3.11) | < 0.001 | 2.28 (1.52–3.42) | < 0.001 | 1.92 (1.26–2.92) | 0.002 | 0.005 |
| P for Trend | < 0.001 | < 0.001 | < 0.001 | ||||||
| Cumulative-AIP | |||||||||
| Cumulative-AIP | 3697 | 176 (4.8) | 1.47 (1.24–1.74) | < 0.001 | 1.54 (1.3–1.84) | < 0.001 | 1.51 (1.23–1.86) | < 0.001 | 0.001 |
| Tertile 1 [− 1.194, 0.644] | 1232 | 41 (3.3) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Tertile 2 [0.645, 1.366] | 1232 | 56 (4.5) | 1.38 (0.92–2.09) | 0.122 | 1.43 (0.94–2.16) | 0.092 | 1.32 (0.87–2.01) | 0.195 | 0.203 |
| Tertile 3 [1.367, 4.745] | 1233 | 79 (6.4) | 1.99 (1.35–2.92) | < 0.001 | 2.14 (1.44–3.17) | < 0.001 | 1.79 (1.19–2.69) | 0.005 | 0.011 |
| P for Trend | < 0.001 | < 0.001 | 0.005 | ||||||
| Cumulative AIP-WHtR | |||||||||
| Cumulative AIP-WHtR | 3697 | 176 (4.8) | 2.02 (1.53–2.67) | < 0.001 | 2.21 (1.66–2.94) | < 0.001 | 2.13 (1.51–3.00) | < 0.001 | 0.001 |
| Tertile 1 [− 0.561, 0.334) | 1232 | 39 (3.2) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Tertile 2 [0.334, 0.748) | 1232 | 52 (4.2) | 1.35 (0.88–2.06) | 0.166 | 1.4 (0.91–2.14) | 0.126 | 1.29 (0.84–1.99) | 0.250 | 0.250 |
| Tertile 3 [0.748, 2.796] | 1233 | 85 (6.9) | 2.26 (1.54–3.34) | < 0.001 | 2.48 (1.67–3.69) | < 0.001 | 2.07 (1.37–3.13) | 0.001 | 0.003 |
| P for Trend | < 0.001 | < 0.001 | < 0.001 | ||||||
| Cumulative AIP-WC | |||||||||
| Cumulative AIP-WC | 3697 | 176 (4.8) | 1 (1–1.01) | < 0.001 | 1.01 (1–1.01) | < 0.001 | 1.01 (1.00–1.01) | < 0.001 | 0.001 |
| Tertile 1 [− 87.375, 52.632] | 1232 | 39 (3.2) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Tertile 2 [52.677, 118.310] | 1232 | 54 (4.4) | 1.4 (0.92–2.13) | 0.114 | 1.45 (0.95–2.22) | 0.085 | 1.34 (0.87–2.06) | 0.178 | 0.194 |
| Tertile 3 [118.327, 472.453] | 1233 | 83 (6.7) | 2.21 (1.5–3.26) | < 0.001 | 2.41 (1.62–3.58) | < 0.001 | 2.01 (1.33–3.04) | 0.001 | 0.003 |
| P for Trend | < 0.001 | < 0.001 | 0.001 | ||||||
| Cumulative AIP-BMI | |||||||||
| Cumulative AIP-BMI | 3697 | 176 (4.8) | 1.01 (1.01–1.02) | < 0.001 | 1.02 (1.01–1.02) | < 0.001 | 1.02 (1.01–1.02) | < 0.001 | 0.001 |
| Tertile 1 [− 23.035, 14.622] | 1232 | 40 (3.2) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Tertile 2 [14.628, 33.145] | 1232 | 56 (4.5) | 1.42 (0.94–2.15) | 0.097 | 1.5 (0.99–2.27) | 0.058 | 1.38 (0.91–2.12) | 0.133 | 0.152 |
| Tertile 3 [33.164, 1229.961] | 1233 | 80 (6.5) | 2.07 (1.4–3.05) | < 0.001 | 2.32 (1.56–3.45) | < 0.001 | 1.92 (1.27–2.89) | 0.002 | 0.005 |
| P for Trend | < 0.001 | < 0.001 | 0.002 | ||||||
| Change in the AIP | |||||||||
| Class 1 | 1345 | 44 (3.3) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Class 2 | 1649 | 79 (4.8) | 1.49 (1.02–2.17) | 0.038 | 1.55 (1.06–2.28) | 0.023 | 1.38 (0.94–2.04) | 0.100 | 0.126 |
| Class 3 | 703 | 53 (7.5) | 2.41 (1.6–3.64) | < 0.001 | 2.68 (1.76–4.08) | < 0.001 | 2.28 (1.46–3.55) | < 0.001 | 0.003 |
| P for Trend | < 0.001 | < 0.001 | < 0.001 | ||||||
| Change in the AIP-WHtR | |||||||||
| Class 1 | 1588 | 50 (3.1) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Class 2 | 1547 | 83 (5.4) | 1.74 (1.22–2.49) | 0.002 | 1.83 (1.27–2.63) | 0.001 | 1.63 (1.13–2.36) | 0.010 | 0.017 |
| Class 3 | 562 | 43 (7.7) | 2.55 (1.68–3.88) | < 0.001 | 2.87 (1.86–4.41) | < 0.001 | 2.39 (1.50–3.79) | < 0.001 | 0.003 |
| P for Trend | < 0.001 | < 0.001 | < 0.001 | ||||||
| Change in the AIP-WC | |||||||||
| Class 1 | 1624 | 51 (3.1) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Class 2 | 1547 | 83 (5.4) | 1.75 (1.22–2.5) | 0.002 | 1.85 (1.29–2.66) | 0.001 | 1.64 (1.14–2.37) | 0.008 | 0.016 |
| Class 3 | 526 | 42 (8) | 2.68 (1.76–4.08) | < 0.001 | 3.01 (1.96–4.63) | < 0.001 | 2.56 (1.61–4.07) | < 0.001 | 0.003 |
| P for Trend | < 0.001 | < 0.001 | < 0.001 | ||||||
| Change in the AIP-BMI | |||||||||
| Class 1 | 1621 | 53 (3.3) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Class 2 | 1527 | 84 (5.5) | 1.72 (1.21–2.45) | 0.002 | 1.84 (1.29–2.64) | 0.001 | 1.62 (1.13–2.34) | 0.009 | 0.017 |
| Class 3 | 549 | 39 (7.1) | 2.26 (1.48–3.46) | < 0.001 | 2.67 (1.72–4.13) | < 0.001 | 2.22 (1.38–3.56) | 0.001 | 0.003 |
| P for Trend | < 0.001 | < 0.001 | < 0.001 | ||||||
Model 1: no covariates were adjusted. Total data sample size 3697
Model 2: adjusted for age, sex, marital status, residence, educational level, drinking status, smoking status. Total data sample size equaled 3697, used sample size equaled 3696
Model 3: adjusted for age, sex, marital status, residence, educational level, drinking status, smoking status, SBP, DBP, TC2012, LDL2012, eGFR and diabetes. Total data sample size equaled 3697, used sample size equaled 3651
AIP atherogenic index of plasma; WHtR waist-to-height ratio; WC waist circumference; BMI body mass index; SBP systolic blood pressure; DBP diastolic blood pressure; TC total cholesterol; LDL low-density lipoprotein cholesterol; eGFR: estimated glomerular filtration ratio
*P value adjusted using the False Discovery Rate (FDR) correction. if P value < 0.001, we set the P value to 0.001 for FDR correction
The study assessed the performance and interrelationships of AIP and its modified indices in predicting stroke risk through ROC curve analysis, Delong's test of the P-value matrix, and IDI analysis. ROC curve analysis: AIP-WHtR showed the highest AUC value of 0.600 (95% CI 0.559–0.642), slightly outperforming AIP-WC (0.598[95% CI 0.557–0.640]), AIP-BMI (0.592[95% CI 0.551–0.633]), and AIP (0.586[95% CI 0.545–0.627]) (Fig.S2 A-D). P-value matrix: AIP-WHtR exhibited stronger predictive ability than AIP (P < 0.001) and AIP-BMI (P = 0.001), but there was no statistically difference compared to AIP-WC (P = 0.32) (Fig.S2 E). Similarly, the ROC curve analysis for cumulative AIP and its modified indices demonstrated comparable findings (Fig.S3 A-E). IDI Analysis: Compared to AIP, all modified indices showed improved predictive performance. AIP-WHtR demonstrated the best performance with an IDI value of 0.055 (95% CI 0.030–0.080) and a P value of < 0.001. AIP-WC followed with an IDI value of 0.051 (95% CI 0.027–0.073), while AIP-BMI showed more limited improvement with an IDI value of 0.033 (95% CI 0.007–0.060) (Fig. 3A). Likewise, the IDI analysis of cumulative AIP and its modified indices yielded comparable findings (Fig. 3B).
Fig. 3.
Comparison of predictive performance between AIP and its modified indices using IDI. A IDI values comparing the predictive ability of baseline AIP-WC, AIP-WHtR, and AIP-BMI versus baseline AIP, and versus each other. B IDI values comparing the predictive ability of cumulative AIP-WC, cumulative AIP-WHtR, and cumulative AIP-BMI versus cumulative AIP, and versus each other. Squares represent the IDI point estimate, and horizontal lines represent the 95% confidence intervals (CI). An IDI value significantly greater than 0 (P < 0.05) indicates improved prediction discrimination by the first index listed in the comparison relative to the second. AIP: atherogenic index of plasma; WC: waist circumference; WHtR: waist-to-height ratio; BMI: body mass index; IDI: integrated discrimination improvement; CI confidence interval
The RCS model analysis evaluated the association between AIP and its modified indices and incident stroke, with all models adjusted for potential confounders (Fig. 4). Key findings include: 1. Significant associations: All AIP and its modified indices showed significant associations incident stroke (P < 0.001). 2. Non-linear relationships: AIP and AIP-BMI displayed non-linear relationships with incident stroke (P = 0.021 and P = 0.040). AIP-WC and AIP-WHtR primarily showed linear relationships (P = 0.066 and P = 0.103). 3. Cumulative AIP and its modified indices: Cumulative AIP, AIP-WC, AIP-WHtR, and AIP-BMI were all significantly associated with incident stroke (P < 0.001). Non-linear relationship tests for cumulative AIP and its modified indices were non-significant (P > 0.05). When analyzing the progression to advanced CKM directly, baseline AIP and its modified indices showed a more pronounced non-linear effect, while cumulative/change indices maintained a graded association (Table S8).
Fig. 4.
Dose–response relationship between baseline/cumulative AIP and its modified indices and stroke risk using RCS. This figure shows RCS curves depicting the association between incident stroke risk and baseline (A, blue)/cumulative (B, red) AIP, and AIP modified by WC, WHtR, or BMI. Curves represent OR adjusted for age, sex, marital status, residence, educational level, drinking status, smoking status, diabetes, SBP, DBP, TC, LDL, and eGFR, with 95% CI shaded. All baseline and cumulative indices were significantly associated with stroke risk (P < 0.001). Notably, baseline AIP and AIP-BMI showed significant non-linear associations (P < 0.05), while baseline AIP-WC and AIP-WHtR, along with all cumulative indices, displayed predominantly linear associations (P for non-linearity > 0.05 for cumulative). The observed linearity, particularly for baseline AIP-WHtR and AIP-WC, may simplify clinical interpretation. RCS restricted cubic spline; AIP atherogenic index of plasma; WC waist circumference; WHtR waist-to-height ratio; BMI, body mass index; OR, odds ratio; CI, confidence interval; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; LDL, low-density lipoprotein cholesterol; eGFR, estimated glomerular filtration rate
Tables S9-S16 provide detailed associations of AIP and its modified indices, their changes, and cumulative measures with incident stroke stratified by risk factors. After fully adjusting for potential confounders, no significant interaction was observed between AIP and its modified indices and the subgroup variables, except for a marginal sex-specific interaction with cumulative AIP (p for interaction = 0.048, Table S13). Subgroup analysis showed that there was no significant statistical correlation between AIP and its modified indices (AIP-WC, AIP-WHtR, AIP-BMI) and incident stroke in early CKM (stages 0–1) populations (P > 0.05). In contrast, in advanced CKM (stages 2–3) populations, the associations between AIP and its modified indices and incident stroke were amplified (P < 0.001). Furthermore, analysis focusing specifically on the risk of progressing to advanced CKM as the outcome revealed distinct dose–response patterns (Table S8): baseline levels of AIP and its modified indices exhibited a non-linear effect (risk primarily elevated in the highest tertile), whereas cumulative exposure and changes in these indices showed a more graded increase in risk across tertiles/classes.
WQS analyses
The WQS analyses (Fig. 5) were conducted using the better-performing comprehensive indicator, specifically AIP-WHtR. The results of the WQS regression analyses show that among the variables assessed, TG had the highest relative contribution weights to incident stroke in 2012 and 2015, with weights of 0.484 and 0.618, respectively, suggesting that TG plays an important role in modifying stroke risk over time.
Fig. 5.
Relative weights of HDL, WHtR, and TG contributing to stroke risk via WQS regression (2012 vs. 2015). This figure shows positive mean weights for HDL, WHtR, and TG derived from a WQS regression analysis of AIP-WHtR. The analysis, adjusted for covariates (age, sex, marital status, residence, education, drinking, smoking, diabetes, SBP, DBP, TC, LDL, eGFR) using 1000 bootstrap iterations, quantifies the contribution of each component to stroke risk. Panel A (2012) and Panel B (2015) display these weights. TG consistently showed the highest weight (0.484 in 2012, 0.618 in 2015), indicating its primary role in the index's association with stroke risk. WQS, weighted quantile sum; AIP, atherogenic index of plasma; WHtR, waist-to-height ratio; HDL, high-density lipoprotein; TG, triglycerides; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; LDL, low-density lipoprotein; eGFR, estimated glomerular filtration rate
Mediation analysis
Mediation analysis (Table 3) revealed that the ePWV played a mediating role in the association between AIP-WHtR and incident stroke. After adjusting for multiple covariates, including sex, marital status, residence, educational level, drinking, smoking, diabetes, TC, LDL, and eGFR, the proportion of the ePWV-mediated effect of AIP-WHtR on incident stroke was 6.48% (P = 0.015).
Table 3.
Mediation analysis of the effect of ePWV on the relationship between AIP-WHtR and stroke
| Independent variable | Mediator | Total effect Coeffcient (95%CI) | P value | Indirect effect Coeffcient (95%CI) | P value | Direct effect Coeffcient (95%CI) | P value | Proportion mediated (%) |
|---|---|---|---|---|---|---|---|---|
| AIP-WHtR | ePWV | 0.154 (0.062, 0.273) | < 0.001 | 0.010 (0.003, 0.019) | < 0.001 | 0.144 (0.056, 0.261) | < 0.001 | 6.48 |
The model was adjusted for multiple variables, including sex, marital status, residence, educational level, drinking, smoking, diabetes, TC, LDL, and eGFR. Age, SBP, and DBP were excluded in the multivariate analysis due to they were included in the formula of ePWV
AIP atherogenic index of plasma; WHtR waist-to-height ratio; ePWV estimated pulse wave velocity; TC total cholesterol; LDL low-density lipoprotein cholesterol; eGFR estimated glomerular filtration ratio; SBP systolic blood pressure; DBP diastolic blood pressure
Discussion
In this large-scale nationwide cohort study of Chinese adults aged 45 and older with CKM stages 0–3, our findings demonstrate that both AIP and its modified indices are significantly associated with incident stroke. Compared to AIP alone, modified AIP indices, particularly AIP-WHtR, exhibit a stronger correlation and superior predictive performance for incident stroke. This underscores the importance of dynamically monitoring AIP, its modified indices (especially AIP-WHtR), and their cumulative effects in assessing incident stroke. Subgroup analyses indicated that this association was stable and consistent across different subgroups (p for interaction > 0.05). Mediation analysis revealed ePWV as a key mediator linking AIP-WHtR and incident stroke, suggesting potential mechanisms via the metabolic-vascular pathway. Finally, WQS regression analysis identified TG as the variable with the highest relative contribution to the AIP-WHtR-associated stroke risk model in both 2012 and 2015.
The most common cause of CKM syndrome is an excess of adipose tissue, its dysfunction, or both [1]. When adipose tissue, particularly visceral fat, becomes dysfunctional, pro-inflammatory and pro-oxidative factors increase, leading to damage in arterial, cardiac, and renal tissues, thereby promoting the occurrence of stroke [36]. A review by The Lancet Diabetes & Endocrinology has suggested that measuring TG concentrations is essential in the assessment of visceral fat, particularly through metrics such as WC, WHtR, and BMI [37]. An increasing body of evidence indicates that incorporating obesity-related indicators, such as BMI, WC, and WHtR, provides a more comprehensive evaluation of stroke and metabolic risks [12–17]. AIP, as an emerging biomarker, has been proposed for assessing dyslipidemia and predicting the risk of insulin resistance (IR) [38]. A substantial amount of research confirms that IR is an independent risk factor for stroke [39]. Additionally, the majority of stroke events are attributed to atherosclerosis, a process significantly promoted by abnormal lipid metabolism. From a pathophysiological perspective, TG-rich remnant particles have been shown to contribute to the formation and progression of atherosclerotic plaques [1]. Furthermore, beyond its role in reverse cholesterol transport, HDL particles exhibit a wide range of beneficial biological activities, including anti-atherosclerotic effects [40]. Our study demonstrates that in CKM stage 0–3 populations, AIP and its modified indices are significantly associated with incident stroke. Compared to AIP alone, the modified indices of AIP show a stronger correlation with incident stroke and superior predictive ability for stroke occurrence. Among these, AIP-WHtR stands out as the most prominent index, demonstrating both robust statistical associations and enhanced predictive performance. This study emphasizes the importance of dynamically monitoring AIP and its modified indices, as well as their cumulative effects, on incident stroke. Specifically, the study highlights the critical role of AIP-WHtR in this context, underscoring the need for comprehensive and dynamic monitoring of these indices. Therefore, in CKM stage 0–3 patients, AIP and its modified indices, as cost-effective and easy-to-use measures, may serve as effective predictors of stroke development and progression.
Our RCS analyses indicated mainly linear associations, but the relationship between AIP/AIP-BMI and incident stroke was nonlinear, suggesting complexity. The RCS curves reflect the shape of the association, not data dispersion, and are influenced by sample size, disease factors, and modeling choices [14].
Our subgroup analysis reveals that AIP and its modified indices show no significant statistical correlation with incident stroke in early CKM populations. This may be attributed to the fact that these individuals lack significant metabolic or stroke risk factors or have not yet reached a substantive threshold. In contrast, in advanced CKM populations with higher metabolic risk factors, the associations between AIP and its modified indices and incident stroke are amplified. This amplification aligns with findings regarding CKM progression itself; analysis of incident advanced CKM as an outcome indicated that while baseline AIP and its modified indices operate via a non-linear threshold effect in these higher-risk individuals, the risk associated with cumulative exposure or changes over time follows a more graded pattern. This suggests that while a high baseline level might signify crossing a critical risk threshold for CKM advancement, ongoing exposure or worsening profiles contribute progressively to risk. This distinction between baseline threshold and cumulative/dynamic graded effects within advanced CKM stages highlights the importance of considering both static and longitudinal assessments. No significant differences in the relationships between AIP and its modified indices and incident stroke were observed across all subgroups, with the sole exception of a marginally significant interaction between cumulative AIP and sex (P for interaction = 0.048). This sex-specific phenomenon may suggest that differences in hormonal levels or metabolic pathways between sexes play a role in the pathophysiological mechanisms of AIP. Nevertheless, the overall association between elevated AIP and its modified indices and an increased risk of stroke remained stable across the various subgroups within the study population, indicating that our findings may have general applicability to broader populations.
WQS analysis revealed that TG has a relatively high weight in AIP-WHtR regulation, suggesting that controlling TG levels is crucial for managing AIP-WHtR. Based on this finding, we recommend establishing TG monitoring as a primary intervention target, emphasizing stringent TG control within nutritional protocols, and prioritizing the initiation of pharmacological treatment for patients presenting with elevated TG levels.We uniquely assessed the mediating role of ePWV and found that it accounted for 6.48% of the total effect of AIP-WHtR on stroke risk. Arterial stiffness reflects vascular aging and loss of elasticity, which are well-established contributing factors for stroke [32, 33]. Age and blood pressure are the two primary determinants of arterial stiffness [41]. Interestingly, emerging evidence convincingly demonstrates a strong correlation between ePWV, calculated based on age and blood pressure, and arterial stiffness [34]. Multiple studies have confirmed that AIP and WHtR are closely associated with increased arterial stiffness [42–44]. However, it remained unclear whether arterial stiffness mediates the relationship between AIP-WHtR and incident stroke. Our study reveals that ePWV, as a reflection of arterial stiffness, serves as a mediating factor in the association between elevated AIP-WHtR and increased stroke risk. The implication of this result is that clinical strategies should center on direct pathogenic pathways, supplemented by the monitoring of vascular stiffness markers for enhanced risk assessment.
This nationwide cohort study offers several significant highlights. It is the first to demonstrate that combining AIP with obesity-related indices, particularly AIP-WHtR, yields stronger statistical correlations and superior predictive performance for stroke compared to AIP alone. Methodological strengths include using IDI analysis alongside AUC for more reliable predictive comparisons, analyzing results robustly across value, change, and cumulative effects, and employing both WQS and mediation analyses for deeper mechanistic insights. However, the study acknowledges limitations: the long follow-up intervals and limited repeated measurements in the CHARLS database hinder the distinction between short-term and cumulative effects; excluding individuals with missing data introduces potential bias, alongside possible residual confounding; the findings' generalizability to other populations is limited; reliance on self-reported CVD outcomes may affect accuracy, necessitating future validation; incident strokes were identified through self-reports, which could introduce inaccuracies due to recall bias and misclassification, thereby affecting the precision of outcome measurements; however, the reliability of self-reported stroke incidents has been supported by their alignment with medical records in other large-scale studies [45, 46].
Conclusions
AIP and its modified indices, especially AIP-WHtR, are significantly associated with stroke risk in CKM stages 0–3. AIP-WHtR demonstrates superior performance in this association, primarily driven by TG and partially mediated by ePWV. Crucially, dynamic monitoring of these indices is vital for precise individual stroke risk assessment. Consequently, AIP and its modified indices, particularly AIP-WHtR, offer a refined tool for stratifying stroke risk in CKM stages 0–3 individuals. This lays a solid foundation for developing personalized, potentially TG-targeting, prevention strategies, underscoring their significant clinical translational potential and application value.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Author contributions
Shuchuan Miao: Conceptualization, Methodology, Data curation, Writing- Original draft preparation, Funding acquisition. Xiaoyan Wang,Pengfei Wen: Visualization, Investigation, Supervision, Writing- Reviewing and Editing. Yun Liao, Tao Wu, Lin Zeng, Yuxing Huang and Xiaomei Song: Data Cleaning. Zhen Xiong, Lisha Deng, Dingjun Li: Study Supervision.
Funding
The authors declare that this study was conducted without any financial support.
Data availability
Find some help on our Data availability statements page.
Declartions
Ethics approval and consent to participate
The studies involving human participants were reviewed and approved by the Ethics Review Committee of Peking University. The patients/participants provided written informed consent to participate in this study.
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.
Xiaoyan Wang and Pengfei Wen contributed equally to this work.
References
- 1.Ndumele CE, Neeland IJ, Tuttle KR, et al. A synopsis of the evidence for the science and clinical management of cardiovascular-kidney-metabolic (CKM) syndrome: a scientific statement from the American Heart Association. Circulation. 2023;148(20):1636–64. 10.1161/CIR.0000000000001186. [DOI] [PubMed] [Google Scholar]
- 2.Ndumele CE, Rangaswami J, Chow SL, et al. Cardiovascular-kidney-metabolic health: a presidential advisory from the American Heart Association. Circulation. 2023;148(20):1606–35. 10.1161/CIR.0000000000001184. [DOI] [PubMed] [Google Scholar]
- 3.Roth GA, Mensah GA, Johnson CO, et al. Global Burden of Cardiovascular Diseases and Risk Factors, 1990–2019. J Am Coll Cardiol. 2020;76(25):2982–3021. 10.1016/j.jacc.2020.11.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Sidney S, Quesenberry CP, Jaffe MG, et al. Recent trends in cardiovascular mortality in the United States and public health goals. JAMA Cardiol. 2016;1(5):594. 10.1001/jamacardio.2016.1326. [DOI] [PubMed] [Google Scholar]
- 5.Malik S, Wong ND, Franklin SS, et al. Impact of the metabolic syndrome on mortality from coronary heart disease, cardiovascular disease, and all causes in United States adults. Circulation. 2004;110(10):1245–50. 10.1161/01.CIR.0000140677.20606.0E. [DOI] [PubMed] [Google Scholar]
- 6.Khan SS, Coresh J, Pencina MJ, et al. Novel prediction equations for absolute risk assessment of total cardiovascular disease incorporating cardiovascular-kidney-metabolic health: a scientific statement from the American Heart Association. Circulation. 2023;148(24):1982–2004. 10.1161/CIR.0000000000001191. [DOI] [PubMed] [Google Scholar]
- 7.Gallo G, Desideri G, Savoia C. Update on obesity and cardiovascular risk: from pathophysiology to clinical management. Nutrients. 2024;16(16):2781. 10.3390/nu16162781. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Meisinger C, Freuer D, Schmitz T, et al. Inflammation biomarkers in acute ischemic stroke according to different etiologies. Euro J Neurol. 2024;31(1):e16006. 10.1111/ene.16006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Fernández-Macías JC, Ochoa-Martínez AC, Varela-Silva JA, Pérez-Maldonado IN. Atherogenic index of plasma: novel predictive biomarker for cardiovascular illnesses. Arch Med Res. 2019;50(5):285–94. 10.1016/j.arcmed.2019.08.009. [DOI] [PubMed] [Google Scholar]
- 10.Althwab SA, Abdulmonem WA, Allemailem KS, Alarifi SA, Hamad EM. Prediction of cardiovascular risk factors and metabolic syndrome in adults from Saudi Arabia using the logarithm of triglyceride/HDL-cholesterol ratio. Int J Health Sci (Qassim). 2024;18(2):50–5. [PMC free article] [PubMed] [Google Scholar]
- 11.Onat A, Can G, Kaya H, Hergenç G. “Atherogenic index of plasma” (log10 triglyceride/high-density lipoprotein−cholesterol) predicts high blood pressure, diabetes, and vascular events. J Clin Lipidol. 2010;4(2):89–98. 10.1016/j.jacl.2010.02.005. [DOI] [PubMed] [Google Scholar]
- 12.Huo RR, Zhai L, Liao Q, You XM. Changes in the triglyceride glucose-body mass index estimate the risk of stroke in middle-aged and older Chinese adults: a nationwide prospective cohort study. Cardiovasc Diabetol. 2023;22(1):254. 10.1186/s12933-023-01983-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Shao Y, Hu H, Li Q, Cao C, Liu D, Han Y. Link between triglyceride-glucose-body mass index and future stroke risk in middle-aged and elderly chinese: a nationwide prospective cohort study. Cardiovasc Diabetol. 2024;23(1):81. 10.1186/s12933-024-02165-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Dang K, Wang X, Hu J, et al. The association between triglyceride-glucose index and its combination with obesity indicators and cardiovascular disease: NHANES 2003–2018. Cardiovasc Diabetol. 2024;23(1):8. 10.1186/s12933-023-02115-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Cui C, Qi Y, Song J, et al. Comparison of triglyceride glucose index and modified triglyceride glucose indices in prediction of cardiovascular diseases in middle aged and older Chinese adults. Cardiovasc Diabetol. 2024;23(1):185. 10.1186/s12933-024-02278-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wang X, Miao S. Enhanced stroke risk stratification using triglyceride-glucose-waist circumference in Chinese adults without diabetes. Am J Prev Med. 2025. 10.1016/j.amepre.2025.01.022. [DOI] [PubMed] [Google Scholar]
- 17.Hong J, Zhang R, Tang H, Wu S, Chen Y, Tan X. Comparison of triglyceride glucose index and modified triglyceride glucose indices in predicting cardiovascular diseases incidence among populations with cardiovascular-kidney-metabolic syndrome stages 0–3: a nationwide prospective cohort study. Cardiovasc Diabetol. 2025;24(1):98. 10.1186/s12933-025-02662-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.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(1):61–8. 10.1093/ije/dys203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Elm EV, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. 2007;335(7624):806–8. 10.1136/bmj.39335.541782.AD. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Dobiás̆ová M, Frohlich J. The plasma parameter log (TG/HDL-C) as an atherogenic index: correlation with lipoprotein particle size and esterification rate inapob-lipoprotein-depleted plasma (FERHDL). Clin Biochem. 2001;34(7):583–8. 10.1016/S0009-9120(01)00263-6. [DOI] [PubMed] [Google Scholar]
- 21.Zhang Y, Chen S, Tian X, et al. Association between cumulative atherogenic index of plasma exposure and risk of myocardial infarction in the general population. Cardiovasc Diabetol. 2023;22(1):210. 10.1186/s12933-023-01936-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Li W, Shen C, Kong W, et al. Association between the triglyceride glucose-body mass index and future cardiovascular disease risk in a population with cardiovascular-kidney-metabolic syndrome stage 0–3: a nationwide prospective cohort study. Cardiovasc Diabetol. 2024;23(1):292. 10.1186/s12933-024-02352-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.He D, Wang Z, Li J, et al. Changes in frailty and incident cardiovascular disease in three prospective cohorts. Eur Heart J. 2024;45(12):1058–68. 10.1093/eurheartj/ehad885. [DOI] [PubMed] [Google Scholar]
- 24.Gao K, Cao LF, Ma WZ, et al. Association between sarcopenia and cardiovascular disease among middle-aged and older adults: findings from the China health and retirement longitudinal study. eClinicalMedicine. 2022;44:101264. 10.1016/j.eclinm.2021.101264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Zheng F, Yan L, Zhong B, Yang Z, Xie W. Progression of cognitive decline before and after incident stroke. Neurology. 2019. 10.1212/WNL.0000000000007716. [DOI] [PubMed] [Google Scholar]
- 26.Li H, Zheng D, Li Z, et al. Association of depressive symptoms with incident cardiovascular diseases in middle-aged and older Chinese adults. JAMA Netw Open. 2019;2(12):e1916591. 10.1001/jamanetworkopen.2019.16591. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Ma YC, Zuo L, Chen JH, et al. Modified glomerular filtration rate estimating equation for Chinese patients with chronic kidney disease. J Am Soc Nephrol. 2006;17(10):2937–44. 10.1681/ASN.2006040368. [DOI] [PubMed] [Google Scholar]
- 28.Chau NVV, Ngoc NM, Nguyet LA, et al. An observational study of breakthrough SARS-CoV-2 delta variant infections among vaccinated healthcare workers in Vietnam. eClinicalMedicine. 2021;41:101143. 10.1016/j.eclinm.2021.101143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Singh A, Yadav A, Rana A. K-means with three different distance metrics. IJCA. 2013;67(10):13–7. 10.5120/11430-6785. [Google Scholar]
- 30.Sinaga KP, Yang MS. Unsupervised K-means clustering algorithm. IEEE Access. 2020;8:80716–27. 10.1109/ACCESS.2020.2988796. [Google Scholar]
- 31.Tanner EM, Bornehag CG, Gennings C. Repeated holdout validation for weighted quantile sum regression. MethodsX. 2019;6:2855–60. 10.1016/j.mex.2019.11.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Mitchell GF, Hwang SJ, Vasan RS, et al. Arterial stiffness and cardiovascular events: the Framingham Heart Study. Circulation. 2010;121(4):505–11. 10.1161/CIRCULATIONAHA.109.886655. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Said MA, Eppinga RN, Lipsic E, Verweij N, van der Harst P. Relationship of arterial stiffness index and pulse pressure with cardiovascular disease and mortality. J Am Heart Assoc. 2018;7(2):e007621. 10.1161/JAHA.117.007621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Greve SV, Blicher MK, Kruger R, et al. Estimated carotid-femoral pulse wave velocity has similar predictive value as measured carotid-femoral pulse wave velocity. J Hypertens. 2016;34(7):1279–89. 10.1097/HJH.0000000000000935. [DOI] [PubMed] [Google Scholar]
- 35.Vlachopoulos C, Terentes-Printzios D, Laurent S, et al. Association of estimated pulse wave velocity with survival: a secondary analysis of sprint. JAMA Netw Open. 2019;2(10):e1912831. 10.1001/jamanetworkopen.2019.12831. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Rana MN, Neeland IJ. Adipose tissue inflammation and cardiovascular disease: an update. Curr Diab Rep. 2022;22(1):27–37. 10.1007/s11892-021-01446-9. [DOI] [PubMed] [Google Scholar]
- 37.Neeland IJ, Ross R, Després JP, et al. Visceral and ectopic fat, atherosclerosis, and cardiometabolic disease: a position statement. Lancet Diabetes Endocrinol. 2019;7(9):715–25. 10.1016/S2213-8587(19)30084-1. [DOI] [PubMed] [Google Scholar]
- 38.Qu L, Fang S, Lan Z, et al. Association between atherogenic index of plasma and new-onset stroke in individuals with different glucose metabolism status: insights from CHARLS. Cardiovasc Diabetol. 2024;23(1):215. 10.1186/s12933-024-02314-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Yang Y, Huang X, Wang Y, et al. The impact of triglyceride-glucose index on ischemic stroke: a systematic review and meta-analysis. Cardiovasc Diabetol. 2023;22(1):2. 10.1186/s12933-022-01732-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Rye KA, Bursill CA, Lambert G, Tabet F, Barter PJ. The metabolism and anti-atherogenic properties of HDL. J Lipid Res. 2009;50(Suppl):S195-200. 10.1194/jlr.R800034-JLR200. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Townsend RR, Wilkinson IB, Schiffrin EL, et al. Recommendations for improving and standardizing vascular research on arterial stiffness: a scientific statement from the American Heart Association. Hypertension. 2015;66(3):698–722. 10.1161/HYP.0000000000000033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Fan B, Zhang T, Li S, et al. Differential roles of life-course cumulative burden of cardiovascular risk factors in arterial stiffness and thickness. Can J Cardiol. 2022;38(8):1253–62. 10.1016/j.cjca.2022.03.009. [DOI] [PubMed] [Google Scholar]
- 43.van den Munckhof ICL, Holewijn S, de Graaf J, Rutten JHW. Sex differences in fat distribution influence the association between BMI and arterial stiffness. J Hypertens. 2017;35(6):1219–25. 10.1097/HJH.0000000000001297. [DOI] [PubMed] [Google Scholar]
- 44.Ishida A, Taira H, Shinzato T, Ohya Y. Association between visceral fat mass and arterial stiffness among community-based screening participants. Hypertens Res. 2023;46(11):2488–96. 10.1038/s41440-023-01350-7. [DOI] [PubMed] [Google Scholar]
- 45.Steptoe A, Breeze E, Banks J, Nazroo J. Cohort profile: the English longitudinal study of ageing. Int J Epidemiol. 2013;42(6):1640–8. 10.1093/ije/dys168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Valtorta NK, Kanaan M, Gilbody S, Hanratty B. Loneliness, social isolation and risk of cardiovascular disease in the English longitudinal study of ageing. Eur J Prev Cardiol. 2018;25(13):1387–96. 10.1177/2047487318792696. [DOI] [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
Find some help on our Data availability statements page.