Carotid intima-media thickening is an important indicator of early atherosclerotic changes in arteries. Early identification and active intervention can effectively reverse the condition.

To explore the association between the longitudinal trajectories of non-high-density lipoprotein cholesterol (nHDL-C) and carotid intima-media thickening in adults, and to predict the risk of carotid intima-media thickening in a health checkup population.

This ambidirectional cohort study enrolled individuals who participated in health examinations at Peking University Third Hospital between 2013 and 2023. Baseline data and physical examination indicators were collected, and CIMT was measured. Participants with normal baseline CIMT were followed up until the occurrence of carotid intima-media thickening or loss to follow-up. Separate dynamic trajectories of nHDL-C with age were constructed for male and female subjects. All nHDL-C records before the occurrence of the outcome were included. A joint latent class model (JLCM) was used to identify heterogeneous nHDL-C change trajectories and predict the risk differences of CIMT among different trajectories. The optimal number of latent classes was determined using the Akaike information criterion (AIC) , Bayesian information criterion (BIC) , sample-adjusted Bayesian information criterion (SABIC) , entropy>0.5, and conditional independence assumption (Score Test P>0.05) . Cox models were constructed using baseline nHDL-C values and nHDL-C change trajectories, respectively. The area under the receiver operating characteristic (ROC) curve (AUC) and concordance index (C-index) of each model were compared, and the goodness of fit of the models was tested and evaluated.

A total of 5 741 subjects with normal baseline lipid levels were included, with 2 487 males and 3 254 females. Among male participants, 393 developed carotid intima-media thickening. There were statistically significant differences in follow-up time, age, BMI, systolic blood pressure (SBP) , diastolic blood pressure (DBP) , total cholesterol (TC) , triglycerides (TG) , low-density lipoprotein cholesterol (LDL-C) , nHDL-C, and the proportion of hypertension between those with and without CIMT (P<0.05) . Among female participants, 330 developed carotid intima-media thickening. There were statistically significant differences in follow-up time, age, BMI, SBP, DBP, TC, TG, LDL-C, nHDL-C, and the proportion of hypertension between the two groups (P<0.05) . In the male population, the three-class model had the highest entropy, the smallest BIC and SABIC, and met the conditional independence assumption (Score Test P=0.207 9) , so the three-class model was selected as the best-fitting model. In the female population, the four-class model had little change in entropy, BIC, and SABIC compared with the three-class model and met the conditional independence assumption (Score Test P=0.267 8) , so the four-class model was selected as the best-fitting model. Among the three latent classes of nHDL-C in the male check-up population, Class 1 showed a trajectory curve that first slowly increased and then remained stable at a low level, named the "low-level stable group" , accounting for 83.80%; Class 2 showed a rapid increase, named the "rapidly increasing group" , accounting for 1.09%; Class 3 showed a slow increase, named the "slowly increasing group" , accounting for 15.12%. The rapidly increasing group had the highest risk, followed by the slowly increasing group, and the low-level stable group had the lowest risk. Compared with the low-level stable group, the hazard ratios (HR) of the slowly increasing group and the rapidly increasing group in males were 10.51 (95%CI=7.90-13.98) and 23.25 (95%CI=10.40-51.98) , respectively. Among the four latent classes of nHDL-C in the female check-up population, Class 1 showed a stable low level, named the "low-level stable group" , accounting for 93.09%; Class 2 showed a U-shaped trajectory, named the "low-level stable-increasing group" , accounting for 1.26%; Class 3 had stable lipid levels at a moderate level without significant fluctuations, named the "moderate-level stable group" , accounting for 4.58%; Class 4 showed a rapid increase in lipid levels, named the "rapidly increasing group" , accounting for 1.08%. The rapidly increasing group had the highest risk. Before the age of 40, the risks of the low-level stable group, low-level stable-increasing group, and moderate-level stable group were similar. After the age of 40, the CIMT thickening risk of the moderate-level stable group increased rapidly, and after the age of 50, the risk of the low-level stable-increasing group increased rapidly. Compared with the low-level stable group, the HR of the low-level stable-increasing group, moderate-level stable group, and rapidly increasing group in females were 3.69 (95%CI=2.27-5.99) , 15.48 (95%CI=10.56-22.70) , and 13.93 (95%CI=5.44-35.69) , respectively. The results of model goodness-of-fit tests and evaluations showed that in both male and female populations, compared with the baseline model, the Class model and the Class+nHDL-C model had significantly increased AUC and C-index values at multiple time points.

In the health check-up population, both males and females have different trajectories of nHDL-C levels, and different trajectory categories significantly affect the risk of CIMT. Compared with a single baseline nHDL-C value, trajectory classification can more accurately predict the risk of CIMT thickening. Continuous lipid monitoring is of great significance for individual health management. The risk assessment method combined with trajectory analysis helps to identify high-risk individuals early and provides a basis for individual risk stratification and active intervention.

颈动脉内中膜增厚是反映动脉粥样硬化早期改变的重要指标，及时发现、积极干预可以得到有效逆转。

探索非高密度脂蛋白胆固醇（nHDL-C）的动态变化轨迹与成人颈动脉内中膜增厚的关联，预测健康体检人群颈动脉内中膜增厚的风险。

本研究为一项双向性队列研究，纳入2013—2023年在北京大学第三医院参加健康体检的人群为研究对象，收集患者基线资料、体检指标，测量颈动脉内膜中层厚度（CIMT），对基线CIMT正常的研究对象进行随访，直到出现颈动脉内中膜增厚或失访。分别构建男性和女性研究对象nHDL-C随年龄的变化轨迹。纳入结局发生前的所有nHDL-C记录，使用联合潜在类别模型（JLCM）识别异质性nHDL-C变化轨迹并预测不同轨迹与发生颈动脉内中膜增厚的风险差异。采用赤池信息准则（AIC）、贝叶斯信息准则（BIC）、样本调整信息准则（SABIC）、熵值（Entropy>0.5）和满足条件独立假定（Score Test P>0.05）确定最优潜在类别个数。分别使用基线nHDL-C数值和nHDL-C变化轨迹构建Cox模型，比较各模型的受试者工作特征（ROC）曲线下面积（AUC）、一致性指数（C-index）数值，对模型拟合优度进行检验及评估。

共纳入基线血脂处于正常水平的研究对象5 741人，男2 487人，女3 254人。男性研究对象中发生颈动脉内中膜增厚393例，无颈动脉内中膜增厚和颈动脉内中膜增厚者比较，随访时间、年龄、BMI、收缩压（SBP）、舒张压（DBP）、总胆固醇（TC）、三酰甘油（TG）、低密度脂蛋白胆固醇（LDL-C）、nHDL-C、高血压占比比较，差异有统计学意义（P<0.05）；女性研究对象中发生颈动脉内中膜增厚330例，无颈动脉内中膜增厚和颈动脉内中膜增厚者的随访时间、年龄、BMI、SBP、DBP、TC、TG、LDL-C、nHDL-C、高血压占比比较，差异有统计学意义（P<0.05）。男性人群中，3类别模型的熵值最高，BIC、SABIC最小且满足条件独立假定（Score Test P=0.207 9），最终选择3分类作为拟合效果最佳的模型；女性人群4类别模型的熵值、BIC、SABIC与3类别模型相比变化不大，且满足条件独立假定（Score Test P=0.267 8），最终选择4分类作为拟合效果最佳的模型。体检人群中男性nHDL-C的3个潜在类别中，类别1的轨迹曲线表现为先缓慢上升后平稳维持在较低水平，为"低水平稳定组"，占比83.80%；类别2表现为快速上升，为"快速升高组"，占比1.09%；类别3表现为缓慢上升，为"缓慢升高组"，占比15.12%；快速升高组风险最高，其次是缓慢升高组，低水平稳定组发生风险最低。相比于低水平稳定组，男性缓慢升高组和快速升高组的HR分别为10.51（95%CI=7.90~13.98）和23.25（95%CI=10.40~51.98）。体检人群中女性nHDL-C的4个潜在类别中，类别1的轨迹曲线表现为稳定的低水平，为"低水平稳定组"，占比93.09%；类别2的轨迹曲线呈现为"U型"，为"低水平稳定-升高组"，占比1.26%；类别3血脂稳定在中等水平，无明显波动，为"中等水平稳定组"，占比4.58%；类别4表现为血脂水平的快速增加，为"快速升高组"，占比1.08%。快速升高组的风险最高，40岁以前，低水平稳定组、低水平稳定-升高组、中等水平稳定组的风险接近，40岁以后中等水平稳定组颈动脉内中膜增厚风险快速增加，50岁以后低水平稳定-升高组风险快速增加。对各个亚组内4个类别人群的颈动脉内中膜增厚累积发生风险进行对比，相比于低水平稳定组，女性低水平稳定-升高组、中等水平稳定组和快速升高组的HR分别为3.69（95%CI=2.27~5.99）、15.48（95%CI=10.56~22.70）和13.93（95%CI=5.44~35.69）。模型拟合优度检验及评估结果显示，在男性及女性人群中，与基线模型相比，Class模型、Class+nHDL-C模型在多个时点的AUC和C-index值均明显增加。

健康体检人群中，男性和女性nHDL-C水平均存在不同变化轨迹，不同轨迹类别显著影响颈动脉内中膜增厚的发生风险。相比单一基线nHDL-C值，轨迹分类能更精准地预测颈动脉内中膜增厚风险，连续血脂监测对于个体的健康管理具有重要意义。结合轨迹分析的风险评估方法有助于早期识别高危个体，为个体的风险分层和积极干预提供依据。