天津医药 ›› 2025, Vol. 53 ›› Issue (10): 1043-1047.doi: 10.11958/20251059

• 临床研究 • 上一篇    下一篇

类风湿关节炎患者动脉粥样硬化风险预测模型的构建与验证

吕婧(), 朱方颖, 朱凯, 李云, 杨娜, 温淑云, 仲米倩()   

  1. 山东大学齐鲁医院德州医院中医科(邮编 253000)
  • 收稿日期:2025-03-16 修回日期:2025-07-10 出版日期:2025-10-15 发布日期:2025-10-12
  • 通讯作者: E-mail:391529383@qq.com
  • 作者简介:吕婧(1989),女,主治医师,主要从事风湿免疫疾病方面研究。E-mail:17653486690@163.com
  • 基金资助:
    山东省中医药科技项目(2020M176)

Construction and verification of atherosclerosis risk prediction model for rheumatoid arthritis patients

LYU Jing(), ZHU Fangying, ZHU Kai, LI Yun, YANG Na, WEN Shuyun, ZHONG Miqian()   

  1. Department of Traditional Chinese Medicine, Dezhou Hospital, Qilu Hospital, Shandong University, Dezhou 253000, China
  • Received:2025-03-16 Revised:2025-07-10 Published:2025-10-15 Online:2025-10-12
  • Contact: E-mail:391529383@qq.com

摘要:

目的 基于Lasso-Logistic回归构建类风湿关节炎(RA)患者动脉粥样硬化(AS)风险预测模型,为个体化临床干预提供科学依据。方法 回顾性收集344例RA患者的临床数据,其中单纯RA组258例和RA+AS组86例。比较2组临床特征和入院首次实验室检查指标的差异。采用Lasso回归筛选关键预测变量,并结合Logistic回归构建预测模型。通过受试者工作特征(ROC)曲线和曲线下面积(AUC)评价模型的区分度,Hosmer-Lemeshow检验评估校准度,并使用决策曲线分析验证模型的临床适用性。结果 筛选出RA病程、28项关节疾病活动度(DAS28)评分、C反应蛋白(CRP)、三酰甘油(TG)、高密度脂蛋白胆固醇(HDL-C)、空腹血糖(FBG)水平和高血压7个预测变量。RA患者AS风险预测模型为Logit(P)=-2.674+0.605×RA病程+0.393×DAS28评分+0.310×CRP+1.346×TG-2.289×HDL-C+0.679×FBG+0.711×高血压。构建的模型AUC为0.965(95%CI:0.943~0.987),Hosmer-Lemeshow χ2=0.547,P=1.000,显示模型区分度和校准度较好。临床决策曲线分析显示概率阈值在7%~92%范围内时,模型具有较高的临床实用性。结论 基于Lasso-Logistic回归构建的RA患者AS风险预测模型可有效识别高风险患者,为制定个体化的预防和治疗策略提供支持。

关键词: 关节炎, 类风湿, 动脉粥样硬化, 危险因素, 预测模型, Lasso-Logistic回归

Abstract:

Objective To construct a risk prediction model for atherosclerosis (AS) in patients with rheumatoid arthritis (RA) based on Lasso-Logistic regression analysis and provide a scientific basis for individualized clinical intervention. Methods The retrospective clinical data were collected from 344 RA patients, including 86 patients with AS (RA+AS group) and 258 patients with without AS (RA group). The clinical characteristics and initial laboratory test results were compared between the two groups. Lasso regression was used to screen the key predictive variables, and Logistic regression was combined to construct the prediction mode. The discrimination of the model was evaluated through the receiver operating characteristic (ROC) curve and the area under the curve (AUC). The Hosmer-Lemeshow test was used to assess the calibration, and decision curve analysis was used to verify the clinical applicability of the model. Results Seven predictive variables were identified including RA disease duration, DAS28 score, C-reactive protein (CRP), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), fasting blood glucose (FBG) and hypertension. The risk prediction model for AS in RA patients was: Logit (P)=-2.674+0.605×RA disease duration +0.393×DAS28 score+0.310×CRP+1.346×TG- 2.289×HDL-C+0.679×FBG+0.711×hypertension. The AUC of the model was 0.965 (95% CI: 0.943-0.987), and the Hosmer-Lemeshow test showed χ2=0.547, P=1.000, indicating good discrimination and calibration. Clinical decision curve analysis showed that the probability threshold ranged from 7% to 92%, demonstrating high clinical applicability. Conclusion The AS risk prediction model constructed in this study for RA patients can effectively identify high-risk individuals, supporting the development of personalized prevention and treatment strategies.

Key words: arthritis, rheumatoid, atherosclerosis, risk factors, prediction model, Lasso Logistic regression

中图分类号: