Multiple risk factors predict carotid intima media thickness in a 10 year prospective study

Abstract

Carotid intima‒media thickness (CIMT) is an early marker of atherosclerosis, and identifying its risk factors is crucial for prevention. This study aimed to explore the risk factors for CIMT, develop a risk prediction model, evaluate its predictive performance, and propose prevention strategies. A prospective cohort was established using health check-up data from the Health Management Center of the Third Xiangya Hospital, Central South University. Participants with physiological CIMT values at baseline and complete clinical data were followed from 2012 to 2023, with at least two examinations recorded. Nonlinear relationships were analyzed via restricted cubic splines, whereas CIMT risk was assessed via Kaplan‒Meier curves. Independent risk factors were identified via Cox regression models and incorporated into a risk prediction model. Among 19,212 participants (64.6% males, 35.4% females), 2,087 developed CIMT, with 1,633 showing an increase of more than 0.2 mm. Multivariate Cox regression identified age (> 40 years), LDL upper tertile, hypertension, impaired fasting glucose, and current smoking as independent risk factors, whereas HDL upper tertile and sleep duration over seven hours were protective factors. The prediction model demonstrated good performance, with AUC values of 0.714, 0.717, and 0.757 at 3, 5, and 8 years, respectively. These findings can be used to enhance CIMT risk assessment and may aid in early prevention strategies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-49451-w.

Introduction

Cardiovascular disease (CVD) is the most significant chronic noninfectious disease that threatens human life and health globally. In recent years, China has become one of the countries with the heaviest burden of CVD worldwide, with deaths caused by CVD accounting for more than 40% of the total deaths in the Chinese population¹. Among these, the burden of atherosclerotic cardiovascular disease (ASCVD) is rapidly increasing in the Chinese population and accounts for more than 60% of cardiovascular deaths². The main causes of CVD are atherosclerosis (AS), stenosis, and occlusion of blood vessels that supply organs such as the heart and brain, leading to damage to the structure and function of blood vessels^3,4. As a major cause of CVD, AS primarily affects the intima-media of arteries, especially large- and medium-sized arteries. Some studies have shown that CIMT can be used as an alternative indicator of subclinical AS, reflecting AS to a certain extent⁵. The results of several epidemiological studies have shown that an increase in CIMT is positively associated with the risk of cardiovascular events^6,7. For every 0.1 mm increase in CIMT, the relative risks of myocardial infarction and stroke are 1.15 and 1.18, respectively⁸. Japanese researchers reported that a CIMT greater than 1.1 mm can be used as a predictor of CVD occurrence⁹. In general, the average CIMT value in adults ranges between 650 and 900 μm, and the annual increase in CIMT ranges from 0 to 40 μm¹⁰. In a cross-sectional analysis of data from 2215 patients with type 2 diabetes mellitus, Jingyi Lu et al. reported that the percentage of time in the target glucose range of 3.9–10.0 mmol/L over a 24-hour period (TIR) was correlated with CIMT¹¹. In a population-based case‒control study¹², Gao et al. reported that serum uric acid was associated with CIMT in patients with different fasting glucose metabolic patterns closely associated with serum uric acid, and serum uric acid, in combination with other factors (e.g., age, SBP, FBG), can be used as a specific model to help predict the incidence of CIMT. A population-based STAAB cohort study of individuals aged between 30 and 79 years revealed that dyslipidemia, hypertension, and smoking were associated with CIMT¹³. Changes in CIMT precede the formation of AS plaques and can be effectively reversed with interventions targeting cardiovascular risk factors. Previous studies have focused on the exploration of risk factors, but the extent of the contribution of each risk factor to CIMT is still unclear, which is not conducive to the primary prevention of CVD. Therefore, the present study aims to clarify the contributions of different risk factors to CIMT, which will provide a reference basis for the early and precise prevention and control of CVD.

Methods

Study design and population

This study is a prospective cohort investigation involving participants from annual health check-ups at the Health Management Center of Xiangya Third Hospital of Central South University in Changsha, China¹⁴. It established the Xiangya CIMT Cohort (hereinafter referred to as the “Xiangya Cohort”). From 2012 to 2023, a total of 128,937 health examination records were compiled into an electronic database. The study applied a prospective cohort design with follow-up through 2023 and included participants who met al.l of the following criteria: complete CIMT measurements at four predefined sites (distal right common carotid artery, right common carotid artery bifurcation, distal left common carotid artery, and left common carotid artery bifurcation), at least two health examinations, and a baseline CIMT within the physiological range (no thickening) at the first examination. Among all the records, 34,178 individuals had complete CIMT measurements at the four sites, and 20,750 of these individuals had undergone at least two CIMT assessments prior to the final follow-up. After excluding 1,538 participants with increased CIMT prior to the baseline examination, the final analytical sample comprised 19,212 participants (Fig. 1). The study received approval from the Research Ethics Committee of Xiangya Third Hospital of Central South University (Approval No. R18030), and all participants provided informed consent.

Fig. 1.

Flow chart of subjects in the prospective cohort design study.

CIMT measurements

Experienced medical examiners used the Ge LOGIQ E9 color ultrasound diagnostic device for measurement. The examinee was placed in a supine position, with a pillow placed behind the neck and the head tilted back and to the contralateral side, and the probe was used to measure the CIMT of the distal section of the common carotid artery and the common carotid artery bifurcation via the multipoint method, i.e., the posterior wall of the common carotid artery at a distance of 1 cm below the inflection point at the beginning of the bifurcation and the posterior wall of the bifurcation. A CIMT ≥ 1.0 mm at the distal common carotid artery or ≥ 1.2 mm at the bifurcation of the common carotid artery, with a change in CIMT ≥ 0.2 mm from baseline, was defined as increased CIMT¹⁵.

Data collection

At baseline and at each follow-up visit, participants underwent a standardized clinical examination, and a trained physician or nurse collected data through structured questionnaires and face‒to-face interviews on sex, age, and education level; lifestyle factors, such as smoking, alcohol consumption, diet, exercise, and hours of sleep; and anthropometric indicators, such as height, weight, waist circumference (WT), hip circumference (HIP), and the waist‒to-hip ratio (WT/HIP); and hemodynamic indicators, such as blood pressure; and health status indicators, such as the presence of underlying diseases, including diabetes mellitus (DM), hypertension (HTN), hyperuricemia (HUA), and hyperhomocysteinemia (HHcy). The participants’ blood samples were also analyzed via a fully automated biochemistry analyzer for alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), total protein (TP), and total protein (TP). protein (TP), albumin (ALB), globulin (GLO), urea, creatinine (Cr), uric acid (UA), total cholesterol (TC), triglyceride (triglyceride (TG), high-density lipoprotein (HDL), low-density lipoprotein (LDL), fasting plasma glucose (FPG), glycated hemoglobin (HbA1c), and serum homocysteine (Hcy) levels. Educational level was categorized as high school and below, college and above, and smoking status was categorized as never smoker, current smoker, and former smoker. Alcohol consumption was categorized as nondrinking or drinking. Sleep duration was categorized as less than 5 h, 5–7 h, more than 7 h, or diet-related information.

Statistical analysis

SPSS 26.0 and R Studio 4.3.2 software were used for statistical analysis. Normally distributed continuous variables are expressed as the means ± standard deviations; nonnormally distributed continuous variables are expressed as medians (Ms) and quartiles (P25, P75); differences between two groups of normally distributed continuous variables were analyzed via the independent samples t test, whereas nonnormally distributed continuous variables were analyzed via the rank sum test. Categorical variables are expressed as rates or component ratios, and differences between groups were analyzed via the chi-square test. Rank correlation was used to perform the dependent variable correlation analysis. Continuous variables were categorized into three subgroups according to the tertiles, and differences in cumulative incidence rates between groups were compared via the log-rank test of the Kaplan‒Meier method. A restricted cubic spline combined with Cox regression was used to fit the nonlinear relationship between exposure risk factors and CIMT. Independent risk factors for CIMT were analyzed via single- and multifactorial Cox risk-proportional regression. Schoenfield residuals were used to evaluate whether the model satisfied the risk-proportionality assumption. Cox proportional risk regression was used to construct a predictive model for CIMT risk. The Cox regression model is expressed as follows:

1

where t denotes the time of event occurrence; h(t) represents the risk function at time t determined by m predictive factors (X₁, X₂…, X_m); β₁, β₂…, β_m are the regression coefficients that measure the effect sizes of the predictive factors; exp is the exponential function; and h₀(t) refers to the baseline risk rate.

The survival function based on the Cox model can be expressed as:

2

The function that calculates the probability of wind for an event can be calculated via the following formula:

3

where H(t) denotes the estimated probability of an individual experiencing CIMT at time t, S₀(t) denotes the survival rate assessed at follow-up time t on the basis of the baseline level, 𝛽_i denotes the regression coefficient of the ith predictor, X_i denotes the value of an individual corresponding to the ith predictor, and 𝑋̅𝑖 denotes the value of the ith predictor in the mean value in the population.

Validation of predictive models and evaluation

The study population was divided into a training set (70%) and a validation set (30%), and the predictive performance of the models was internally validated on the validation set. Seven separate models with progressively increasing predictors were built, and the performance of each model was assessed by the area under the ROC curve (AUC). Delong’s test was used to compare the variability between the models to determine whether increasing the number of predictors significantly improved the predictive efficacy of the models. A difference of P < 0.05 was considered statistically significant. A column-line graphical prediction model for the risk of CIMT was constructed via the rms program package¹⁶, and the predictive value of this prediction model was also assessed by plotting calibration curves and time-dependent ROC curves.

Results

Baseline status of the Xiangya CIMT cohort

A total of 19,212 study subjects were included in this study, with an age distribution of 43.83 ± 10.34 years, of which 12,404 (64.6%) were male, with an age distribution of 43.71 ± 10.2 years, and 6,808 (35.4%) were female, with an age distribution of 44.04 ± 10.58 years. By the end of follow-up in August 2023, among 19,212 participants, 2087 developed increased CIMT, 1634 of whom exhibited a significant increase in CIMT (> 0.2 mm). The age distribution of the participants was 50.43 ± 9.88 years, including 1229 males (75.3%), with an age distribution of 49.88 ± 9.87 years, and 404 females (24.7%), with an age distribution of 52.11 ± 9.74 years. The person-years of follow-up from baseline to the development of increased CIMT or the end of the follow-up period (August 31, 2023) were calculated for each study subject. The incidence density of increased CIMT was calculated via the number of individuals who developed increased CIMT during the follow-up period as the numerator and the total number of person-years of observation in the study population as the denominator. The incidence of increased CIMT was 15.6 person-years, the incidence of increased CIMT in males was 17.6 person-years, the incidence of increased CIMT in females was 11.5 person-years (17.6/1000 vs. 11.5/1000), and the t interval was 1.37–1.72 (P < 0.001). The physical examination population was divided into two groups according to whether increased CIMT had occurred at the end of follow-up, and the characteristics of the baseline indicators were compared between the groups (Table 1).

Table 1.

Baseline characterization of CIMT in the Xiangya CIMT cohort.

Gender(man%)	Total (N = 19212)	With incident (N = 1634)	Without incident (N = 17578)	t, χ2	P Value
	12,404(64.6)	1229(75.3)	11,175(63.6)	89.250	< 0.001
age(year)	43.83 ± 10.34	50.43 ± 9.88	43.21 ± 10.16	−28.165	< 0.001
height	165.11 ± 7.81	165.66 ± 7.40	165.06 ± 7.85	−3.047	0.003
weight	66.65 ± 12.00	69.14 ± 11.43	66.42 ± 12.03	−8.975	< 0.001
SBP	122.09 ± 15.00	127.43 ± 15.75	121.59 ± 14.83	−14.265	< 0.001
DBP	76.25 ± 11.24	79.88 ± 11.34	75.91 ± 11.17	−13.410	< 0.001
BMI	24.32 ± 3.23	25.09 ± 3.13	24.25 ± 3.23	−10.130	< 0.001
WT	82.66 ± 9.86	85.7 ± 9.17	82.38 ± 9.87	−13.751	< 0.001
HIP	94.29 ± 5.95	95.2 ± 5.78	94.21 ± 5.96	−6.438	< 0.001
WT/HIP	0.87 ± 0.07	0.9 ± 0.06	0.87 ± 0.07	−16.197	< 0.001
TBIL	14.76 ± 5.59	15.15 ± 5.58	14.72 ± 5.60	−2.996	0.003
TP	73.51 ± 4.23	73.04 ± 4.50	73.55 ± 4.20	4.374	< 0.001
ALB	46.95 ± 2.81	46.52 ± 2.82	46.98 ± 2.81	6.306	< 0.001
GLO	26.56 ± 3.60	26.52 ± 3.77	26.57 ± 3.58	0.482	0.630
Urea	4.72 ± 1.17	4.97 ± 1.22	4.69 ± 1.16	−8.735	< 0.001
Cr	73.35 ± 16.15	75.73 ± 15.73	73.13 ± 16.17	−6.347	< 0.001
UA	341.1 ± 89.27	348.29 ± 89.01	340.4 ± 89.27	−3.418	< 0.001
CHO	5.04 ± 0.95	5.22 ± 1.02	5.02 ± 0.94	−7.659	< 0.001
logTG	0.17 ± 0.27	0.21 ± 0.26	0.17 ± 0.28	−6.058	< 0.001
HDL	1.38 ± 0.34	1.36 ± 0.34	1.38 ± 0.34	2.222	0.026
LDL	2.82 ± 0.81	2.96 ± 0.90	2.81 ± 0.80	−6.620	< 0.001
ALT	23(16, 34)	25(18, 36)	22(13, 34)	−7.653	< 0.001
AST	22(18, 26)	23(19, 28)	22(18, 26)	−4.759	< 0.001
TG	1.4(0.95, 2.17)	1.55(1.06, 2.32)	1.39(0.95, 2.15)	−6.518	< 0.001
Hcy	12.8(11, 14.8)	13.4(11.6, 15.4)	12.7(10.9, 14.8)	−7.531	< 0.001
FPG	5.25(4.92, 5.64)	5.39(5.0, 5.92)	5.24(4.91, 5.62)	−10.564	< 0.001
Education level				9.584	0.002
High school and below	2769(17.5)	295(20.5)	2474(17.2)
College and above	13,035(82.5)	1143(79.5)	11,892(82.8)
Smoking				47.153	< 0.001
Nonsmoker	12,546(70.1)	965(2.5)	11,581(70.9)
Used to smoke	626(3.5)	68(4.4)	558(3.4)
Current smoker	4714(26.4)	511(33.1)	4203(25.7)
Alcohol				21.364	< 0.001
No	11,078(62.2)	873(56.7)	10,205(62.7)
Yes	6738(37.8)	667(43.3)	6071(37.3)
Fish/seafood				14.352	< 0.001
No	2867(16.2)	297(19.6)	2570(15.8)
Yes	14,866(83.8)	1216(80.4)	13,650(84.2)
Coffee				2.640	0.104
No	11,018(62.2)	909(60.2)	10,109(62.4)
Yes	6704(37.8)	601(39.8)	6103(0.376)
Sugary drinks				16.684	< 0.001
No	9041(51)	847(56.1)	8194(50.5)
Yes	8687(49)	664(43.9)	8023(49.5)
Sleep duration				12.539	0.002
< 5 h	1200(6.7)	136(8.8)	1064(6.5)
5–7 h	11,269(63.1)	966(62.6)	10,303(63.1)
7–9 h	5403(30.3)	442(28.6)	4961(30.4)
HTN				174.120	< 0.001
No	15,664(81.5)	1133(69.4)	14,531(82.7)
Yes	3548(18.5)	500(30.6)	3048(17.3)
IFG				112.880	< 0.001
No	16,993(89.1)	1319(81.2)	15,674(89.8)
Yes	2086(10.9)	306(18.8)	1780(10.2)
DM				128.380	< 0.001
No	17,109(89.1)	1317(80.6)	15,792(89.8)
Yes	2103(10.9)	316(19.4)	1787(10.2)
HUA				1.821	0.177
No	15,124(79.2)	1266(77.9)	13,858(79.4)
Yes	3963(20.8)	359(22.1)	3604(20.6)
HHcy				19.017	< 0.001
No	15,277(79.5)	1230(75.3)	14,047(79.9)
Yes	3935(20.5)	403(24.7)	3532(20.1)

Restrictive cubic spline analysis of multiple variables

Restricted cubic spline analysis was used to explore nonlinear relationships between variables and outcomes. The results showed that the risk of CIMT increased progressively in those older than 40 years, with a WT greater than 80 cm, an SBP greater than 120 mm Hg, an LDL greater than 3 mmol/L, and a urea greater than 4.5 mmol/L. In contrast, the risk of CIMT increased rapidly with FPG between 5.2 and 7.0 mmol/L and then leveled off gradually (Fig. 2A-F).

Fig. 2.

Visualization of the results of various risk factors associated with CIMT using constrained cubic splines. (A) The relationship between Age and the risk ratio of CIMT; (B) The relationship between WT and the risk ratio of CIMT; (C) The relationship between SBP and the risk ratio of CIMT; (D) The relationship between FPG and the risk ratio of CIMT; (E) The relationship between LDL and the risk ratio of CIMT; (F) The relationship between Urea and the risk ratio of CIMT. The red shaded area represents the 95% CI, the red solid line represents the HR value, and the 4 nodes of the restriction cubic spline are 5%, 25%, 75%, and 95%.

Single and multifactor Cox regression analysis of CIMT

The anthropometric indicators of gender, age, BMI, WT, WT/HIP, SBP, DBP; laboratory indicators of ALT, AST, TBIL, TP, ALB, GLO, Urea, Cr, UA, CHO, TG, HDL, LDL, FPG, Hcy; and lifestyle exposures, including education level, smoking, drinking, diet, and sleep were used as potential risk factors for CIMT. As the potential risk factors for CIMT, Cox proportional risk regression model was used to perform one-way Cox regression for the above variables, respectively. Risk factors that were statistically significant in the one-way Cox regression analysis (Supplementary Table S1-S3) were included as independent variables in the multifactorial Cox risk regression model for further exploration. To better understand the complex interactions between these variables and to enhance the robustness of the predictive model, Spearman’s correlation analysis was used to assess potential covariance issues between predicted exposure risk factors prior to modeling; variables with Spearman’s correlation coefficients greater than 0.8 were considered to be highly correlated, and independent variables with covariance were included in only one into the model. Subsequently, forward stepwise regression was used to screen the variables to obtain the final risk predictors for CIMT, including age, HDL, SBP, LDL, IFG, sleep duration, and smoking status, and the associations between the predictors and the risk of CIMT are shown in Table 2, and Figure S1.

Table 2.

Multivariate Cox regression analysis of CIMT risk.

Variables	β	Se	Z	HR	95%CI	P Value
age(year)	1.(ref)
> 40	1.066	0.087	12.324	2.904	2.451–3.441	< 0.000
gender	man(ref)
woman	−0.177	0.116	−1.521	0.838	0.667–1.052	0.128
HDL(mmol/L)
< 1.2	1.(ref)
1.2–1.47	−0.165	0.072	−2.303	0.848	0.737–0.976	0.021
> 1.47	−0.307	0.086	−3.589	0.735	0.622–0.870	< 0.001
LDL(mmol/L)
< 2.47	1.(ref)
2.47–3.13	0.188	0.076	2.485	1.206	1.040–1.399	0.013
> 3.13	0.318	0.094	3.373	1.375	1.143–1.654	0.001
HTN	0.324	0.065	4.955	1.382	1.216–1.571	< 0.001
IFG	0.347	0.078	4.463	1.415	1.215–1.649	< 0.001
Sleep duration
< 5 h	1.(ref)
5–7 h	−0.242	0.101	−2.39	0.785	0.644–0.957	0.017
> 7 h	−0.267	0.109	−2.458	0.766	0.619–0.947	0.014
variables		β	Se	Z	HR	95%CI	P Value
Smoking
Nonsmoker		1.(ref)
Used to smoke		0.021	0.143	0.145	1.021	0.772–1.350	0.885
Current smoker		0.173	0.070	2.465	1.188	1.036–1.363	0.014

HDL high density lipoprotein, LDL low density lipoprotein, HTN hypertension, IFG impaired fasting glycemia, β Beta, Se Standard Error, Z Z score, HR Hazard Ratio, 95%CI 95% Confidence Interval, ref reference group.

Kaplan‒Meier survival curves for risk factors associated with CIMT

Kaplan‒Meier survival curves were used to analyze the cumulative risk of CIMT over time in different subgroups. The results showed that age > 40 years, male, LDL > 3.13 mmol/L, HTN, IFG, sleep duration < 5 h, and smoking accompanied the risk of CIMT with a higher cumulative incidence than in the normal group: whereas HDL > 1.47 mmol/L was accompanied by a reduced risk of CIMT. All subgroups passed the log-rank test (P < 0.001) (Fig. 3A-H). In the Kaplan‒Meier analysis, the number of patients at risk at each time point is shown in Supplementary Table S4.

Fig. 3.

Different cutoff values for various risk factors for CIMT. Risk profile of age (A), sex (B), HDL (C), LDL (D), hypertension (E), impaired fasting glucose (F), sleep duration (G), and smoking status (H) versus CIMT.

Construction and validation of a prediction model for CIMT

Cox proportional risk regression was used to construct a risk prediction model for CIMT by combining multiple predictors. The results of analyzing and comparing the predictive efficacy of different predictors are shown in Supplementary Table S5. The results of the analysis showed that the AUCs of age, gender, hypertension, HDL, LDL, IFG, sleep, and smoking for predicting CIMT were 0.696 (0.683–0.708), 0.559 (0.548–0.571), 0.566 (0.554–0.579), and 0.526 (0.511–0.5420), 0.554 (0.538–0.57), 0.554 (0.534–0.555), 0.516 (0.503–0.529), and 0.542 (0.529–0.555), respectively. C-indices were 0.683, 0.545, 0.599, 0.546, 0.561, 0.546, 0.536, and 0.528, respectively.

According to the magnitude of the predictive efficacy of each of the above predictors on the risk of CIMT, they were sequentially included in the model for different risk combinations from high to low to construct a multifactorial risk prediction model for CIMT, and then optimally screened for the best prediction model for CIMT. Model 1 included age and LDL as predictors, and the prediction model AUC was 0.698 (0.685–0.710), and the C index was 0.688 (0.007); Model 2 included age, LDL, and HTN as predictors, and the prediction model AUC was 0.703 (0.690–0.715), and the C index was 0.691 (0.007); Model 3 contains age, LDL, HTN, and IFG as predictors, and the predictive model AUC is 0.706 (0.693–0.718), with a C index of 0.696 (0.007); Model 4 contains age, LDL, HTN, IFG, and HDL as predictors, and the predictive model AUC is 0.708 (0.695–0.720), with a C index of 0.706 (0.007); Model 5 included age, LDL, HTN, IFG, HDL, and gender as predictors, and the predictive model AUC was 0.714 (0.701–0.726), with a C-index of 0.708 (0.007); Model 6 included age, LDL, HTN, IFG, HDL, gender, and smoking as predictors, and the predictive model AUC was 0.713 (0.701–0.726), with a C index of 0.708 (0.007); Model 7 included age, LDL, HTN, IFG, HDL, gender, smoking, and sleep duration as predictors, and the predictive model AUC was 0.714 (0.701–0.727), with a C index of 0.707 (0.007). The results of the predictive efficacy of different prediction models are shown in Supplementary Table S6. Models 1 to 7, the AUC values of the different predictive models were similar in the train and validation set (Fig. 4A-G), and the differences in the AUC values of the predictive models in the train and validation set were not significantly different using DeLong’s test (P > 0.05), which indicated that the predictive models in this study were not overfitted. Comparing Model 5 with Model 4, the AUC value increased by 0.006 and was statistically significant (P < 0.05); comparing Model 6 and Model 7 with Model 5, there was no statistically significant difference in the AUC value (P > 0.05).

Fig. 4.

Validation of predictive models and evaluation. (A-G) The predictive performance of seven CIMT risk prediction models developed in the Xiangya cohort. The AUC values of Model 1 to Model 7 in the training, validation, and overall sample datasets. The red curve represents the ROC curve of the training sample, the green curve represents the ROC curve of the validation sample, and the blue curve represents the ROC curve of the overall sample. H. ROC curves for predicting carotid intima‒media thickness at different follow-up times.

Based on the predictive efficacy of the prediction models, Model 5 was chosen as the risk prediction model for CIMT in this study. Table 3 demonstrates the basic characteristics of the predictors included in this study. Six predictors contained three continuous variables (age, LDL vs. HDL) and three dichotomous variables (gender, whether HTN, and whether IFG). Table 4 contains the regression coefficients and HRs of the six predictors for the risk of developing CIMT. According to Model 5, at the mean level of all covariates, the 10-year baseline survival rate of CIMT is 0.915. Based on formula (4), the predictive model for the risk of CIMT is:

Table 3.

Basic characteristics of the predictors.

Variables	Unit/Assignment	Mean
age	continuous variable (year)	43.82
HDL	continuous variable (mmol/L)	1.38
LDL	continuous variable (mmol/L)	2.82
gender	0: woman, 1: man	0.65
HTN	0: No, 1:Yes	0.19
IFG	0: No, 1: Yes	0.11

HDL, high density lipoprotein; LDL, low density lipoprotein; HTN, hypertension; IFG, impaired fasting glycemia.

Table 4.

Regression coefficients for predictors in the CIMT risk prediction model.

Predictors	β	Se	Wald	HR (95% CI)	P Value
age	0.05	0.002	20.744	1.053(1.047–1.058)	< 0.001
HDL	−0.52	0.090	−5.780	0.594(0.498–0.709)	< 0.001
LDL	0.25	0.031	8.108	1.285(1.210–1.365)	< 0.001
gender	0.33	0.065	5.063	1.390(1.224–1.579)	< 0.001
HTN	0.11	0.060	1.829	1.116(0.992–1.256)	0.067
IFG	0.29	0.069	4.256	1.341(1.172–1.536)	< 0.001

HTN, hypertension; HDL, high-density lipoprotein; LDL, low-density lipoprotein; IFG, impaired fasting glycemia; β, beta; Se, standard error; 95% CI, 95% confidence interval.

4

where

 

Assuming a 60 male with HDL of 2 mmol/L and LDL of 3 mmol/L, the risk of CIMT is 14.9%. If this man also has hypertension and impaired fasting glucose, the risk of CIMT is 21.3%.

CIMT risk column plot prediction model with calibration curves

Based on the six predictors incorporated in Model 5, a column chart of the risk prediction model for CIMT is presented using the rms package of R Studio (Supplementary Figure S2). The column chart is constructed by assigning an initial scale score of 0–100 to each of the six predictors for each of the six variables, which is then summed to produce a total score, and finally converted to an individual probability (0–100%). Based on the cumulative independent risk factor numerical scores, a higher total score corresponds to a lower survival rate and a higher risk of CIMT. Based on model 5, with 1000 Bootstrap, we plotted calibration curves for predicted survival probabilities at 3, 5 and 8 years of follow-up¹⁷. CIMT at 3, 5 and 8 years exhibits good agreement between predicted and actual survival probabilities, indicating that our model is not overfitted (Supplementary Figure S3). To further analyze the predictive stability of the model at different time points, we plot time-dependent ROC curves, and the AUC of the 3-year, 5-year, and 8-year CIMT risk prediction models are 0.714, 0.717, and 0.757, respectively, and the AUC area is greater than 0.7 at each time point, and the results show that the predictive ability of the model remains stable at different time points (Supplementary Figure S4).

Discussion

Significance and application value of CIMT risk prediction models

Atherosclerosis is a systemic disease that affects multiple blood vessels and organs. In the physiological state, the carotid artery vessel wall consists of three layers: intima, mid-membrane, and adventitia. The earliest observable pathological change is the appearance of lipid patterns or lipid accumulation in the inner layer of the intima-media. In 2010, the American Heart Association recommended carotid ultrasound cardiovascular disease screening in asymptomatic adults¹⁸. Studies have shown¹⁹ that CIMT and its plaque grading or detection rate are positively correlated with the occurrence of coronary artery disease and the extent of lesions. Noninvasive measurement of CIMT by color Doppler ultrasound is not only economical and reproducible but also easy to perform because of the superficial and fixed position of the carotid artery. However, there are fewer predictive models for CIMT in existing studies, and most studies have used CIMT as a predictor to create predictive models related to cardiovascular disease. For example, the Framingham Heart Study (Framingham study) used increased common carotid artery IMT, plaque, and Framingham Risk Score as independent determinants of stroke risk²⁰. The Kuoppio Ischemic Heart Disease Study demonstrated, for the first time, that CIMT was associated with future coronary events, and that for every 0.1 mm increase in CIMT, the risk of developing a follow-up myocardial infarction increased the risk of myocardial infarction by 11%²¹. The present study explores the risk factors for CIMT and establishes a risk prediction model for CIMT, which may enable the prevention of cardiovascular disease to be moved forward. The strengths of this study are: a prospective cohort study based on a large physical examination population with 10 years of follow-up, which has excluded those with CIMT at baseline, has the advantage of long-term follow-up, observation of causality, and exclusion of reverse causality, and is particularly good at studying chronic disease development and the long-term effects of risk factors. For conditions such as CIMT that change slowly, prospective studies can better capture their evolution and avoid changes that may not be detectable with brief observations. By monitoring and studying large populations over a long period of time, studying carotid intima-media thickness in a general medical examination population can help in the early identification of individuals at potential risk of cardiovascular disease. Larger carotid intima-media thickness may suggest early development of atherosclerosis and other cardiovascular diseases even in the absence of obvious symptoms, and by creating a predictive model of CIMT risk, physicians can better individualize medical decisions. The risk of CIMT may vary from patient to patient depending on a variety of factors, including age, gender, biochemical markers, lifestyle, etc. Risk modeling can help physicians more comprehensively assess a patient’s overall risk, and thus better select appropriate treatment options.

Risk factors for CIMT

Independent risk factors for CIMT found in this study included age, SBP, LDL, IFG, sleep duration, and smoking. Previous studies have shown a strong correlation between age and CIMT²². In this study, 19,212 cases of the general population participating in health checkups were included for statistical analysis, with an age range of 18–88 years and a mean age of 43.83 ± 10.34. Kaplan‒Meier survival curve analysis grouped the checkup population by age at intervals of 20 years, which revealed that the risk of the carotid internal intima-media thickness increased continuously with age with the increase of age. In the Cox regression analysis model, the risk of CIMT in subjects older than 40 years of age was 3.635 times higher than that in subjects younger than 40 years of age. Therefore, we should pay high attention to carotid ultrasonography in middle-aged and elderly people over 40 years of age for early detection of individuals at high risk of cardiovascular disease.

In our study population, the mean age at baseline was 43 years for men and 44 years for women, and the incidence of CIMT was 17.6 per 1,000 person-years in men and 11.5 per 1,000 person-years in women (17.6/1,000 vs. 11.5/1,000), which is consistent with previous reports of a higher prevalence of CIMT in men compared with women, who have a lower age at baseline. The higher incidence is consistent with the previously reported 10–15 year lag in coronary artery disease in women compared to men, and this delay may be related to the protective effect of estrogen against coronary atherosclerosis²³. On the other hand, men have relatively high work stress compared with women, and poor lifestyle habits such as smoking, alcohol abuse, and socializing may be responsible for the higher rate of CIMT.

In addition, we found that the percentage of the CIMT group with baseline hypertension and impaired fasting glucose was significantly higher than that of the group without CIMT, with hypertension (30.6% vs. 17.3%), and impaired fasting glucose (18.8% vs. 10.2%), which suggests that both hypertension and impaired fasting glucose are risk factors for CIMT. The results of the present study showed that the effect of FPG on CIMT increased rapidly and then leveled off; therefore, the final risk factor to be included in the model was selected at the stage of impaired fasting glucose rather than at the stage where DM had been diagnosed. A study on the effect of metabolic syndrome on carotid atherosclerosis in the Japanese general population showed a positive correlation between SBP and CIMT, and among the components of metabolic syndrome, hypertension had the strongest correlation with CIMT, suggesting that hypertension has an important role in carotid atherosclerosis formation²⁴. Femia et al. followed up 1,536 study participants over a period of 3.5 years, found that CIMT was 1.45 times higher in hypertensive patients than in the normal population²⁵. Brohall found that CIMT in patients with type 2 diabetes mellitus was, on average, 0.13 mm thicker than that of controls. Several trials have confirmed that the blood glucose level at 120 min after loading is the most significant and independent determinant of CIMT²⁶. Unfortunately, our researchers did not include the recording of blood glucose 2 h after meals, which is one of the limitations of our study. Beyond, we found that sleep duration and smoking history were strongly associated with CIMT. Multifactorial Cox regression analysis in this study found that there was no statistically significant difference in ever smoking compared to nonsmoking (P = 0.885), which may be due to the small sample size of the ever-smoking segment compared to the number of nonsmokers; there was a statistically significant difference in current smoking compared to nonsmoking (P = 0.014), with an 18% increase in the risk of CIMT. Iwasaki²⁷ in Tochigi, Japan found that current smoking and hypertension were the most important factors for increased CIMT. This is consistent with our findings and suggests the importance of anti-smoking education. Sleeping 5–7 h and greater than 7 h were associated with an 11% and 13% reduction in the risk of CIMT, respectively, compared with sleeping less than 5 h. Jade found a weak negative correlation between sleep duration and CIMT in adolescents aged 10–18 years in a cross-sectional study²⁸. Most studies have shown an association between sleep duration and CIMT, which usually follows a U-shaped curve, with elevated CIMT with too little or too much sleep^29,30. Subjects with an average sleep duration of 7–8 h had the lowest CIMT values, but it increased with shorter sleep duration and even with longer sleep duration The shorter or longer the sleep duration, the subjects with 11–12 h of sleep had still higher CIMT values (as compared to 8 h of sleep) elevated CIMT³¹.

Conclusion

In summary, in a large prospective cohort study involving 19,212 participants, we conducted a 10-year follow-up of baseline physiological CIMT and identified independent risk factors associated with subsequent progression. A CIMT risk prediction model was established using these factors. These findings contribute to a deeper understanding of early vascular changes and may provide a basis for risk stratification and preventive strategies to limit the progression of atherosclerosis. Future interventional studies are needed to validate these observations and assess the impact of targeted preventive measures on CIMT progression.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ALB

Albumin

ALT

Alanine aminotransferase

AST

Aspartate transaminase

AUC

Area under curve

AS

Atherosclerosis

BMI

Body mass index

CAD

Coronary artery disease

CIMT

Carotid intima-media thickness

Cr

Creatinine

CHO

Cholesterol

DM

Diabetes mellitus

DBP

Diastolic blood pressure

FPG

Fasting plasma glucose

GLO

Globulin

HbA1c

Glycated hemoglobin

Hcy

Homocysteine

HHcy

Hyperhomocysteinemia

HDL

High density lipoprotein

HIP

Hip circumference

IFG

Impaired fasting glucose

LDL

Low density lipoprotein

ROC

Receiver operating characteristic curve

SBP

Systolic blood pressure

TG

Triglyceride

TP

Total protein

TBIL

Total bilirubin

UA

Uric acid

WT

Waist circumference

WT/HIP

Waist‒to-hip ratio

Author contributions

Q.Y. designed the study, performed analyses, and wrote the manuscript. W-D. L. designed the study, edited the manuscript and supervised the study. K.C and Y.X. collected clinical data and performed analyses, Q.Y., A.Z, Y.W., and F.W. performed the analyses, J.W. collected clinical data, made comments on data analyses and the manuscript. Y.W. edited the manuscript. All authors reviewed the manuscript.

Funding

This work was supported by the Key Program of Regional Innovative Development Joint Funds of the National Natural Science Foundation of China (Grant No. U24A20774), the Chinese Cardiovascular Association-ASCVD Fund (2023-CCA-ASCVD-018), and the Project of State Key Clinical Department (Z2023058).

Data availability

All the data generated or analyzed during this study are included in this published article and in the supplementary materials.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Patients were recruited from Xiangya Third Hospital of Central South University. This study was approved by the Institutional Ethical Review Board (IRB) of the Research Ethics Committee of Xiangya Third Hospital of Central South University (Approval No. R18030). All the subjects signed informed consent forms. All clinical investigations were conducted according to the principles of the Declaration of Helsinki.