中国麻风皮肤病杂志 ›› 2026, Vol. 42 ›› Issue (1): 30-34.doi: 10.12144/zgmfskin202601030

• 论著 • 上一篇    下一篇

基于血清TARC、CCL19、LRG1水平构建寻常型银屑病复发预警模型

李新1,王君2,苗国英1,张建忠1,李玲玉1   

  1. 1河北工程大学附属医院皮肤科，河北邯郸，056002； 2邯郸科技职业学院实训中心，河北邯郸，056046
  • 出版日期:2026-01-15 发布日期:2026-01-15

Development of a relapse prediction model for psoriasis vulgaris based on serum TARC, CCL19 and LRG1 levels

LI Xin1, WANG Jun2, MIAO Guoying1, ZHANG Jianzhong1, LI Lingyu1   

  1. 1 Department of Dermatology, Affiliated Hospital of Hebei University of Engineering, Handan 056002, China; 2 Practical Training Center, Handan Vocational College of Science and Technology, Handan 056046, China
  • Online:2026-01-15 Published:2026-01-15

PDF (PC)

3

摘要： 目的：明确寻常型银屑病（psoriasis vulgaris，PV）复发因素，构建预警模型。方法：选取2022年1月至2023年12月河北工程大学附属医院258例PV患者作为研究对象，将样本按照7∶3比例随机分为训练集和测试集，采用门诊或电话形式随访1年，根据复发情况分为复发组和未复发组。比较两组临床资料及血清TARC、CCL19、LRG1水平，采用Logistic回归方程分析PV复发的影响因素，构建4种机器学习模型，绘制受试者工作特征曲线（ROC）及曲线下面积（AUC）分析4种机器学习模型预测性能。结果：258例PV患者中有8例失访，120例复发。训练集、测试集中，复发组血清TARC、CCL19、LRG1水平、SAS及SDS评分、睡眠障碍发生率均高于未复发组（均P＜0.05）。Logistic回归方程结果最终确定4个变量（血清TARC、CCL19、LRG1水平、睡眠障碍）是PV复发的影响因素，用于构建机器学习模型。ROC曲线显示，随机森林算法模型在训练集、测试集中AUC均高于深度学习、梯度提升机、广义线性模型（均P＜0.05）。结论：基于血清TARC、CCL19、LRG1水平及睡眠障碍建立的随机森林算法模型可有效预测PV复发风险。

关键词: 寻常型银屑病, 复发, 预警模型, TARC, CCL19, LRG1

Abstract: Objective: To identify the recurrence factors of psoriasis vulgaris (PV) and construct an early warning model. Methods: A total of 258 PV patients admitted to the Affiliated Hospital of Hebei University of Engineering from January 2022 to December 2023 were enrolled as the research subjects. The samples were randomly divided into a training set and a test set at a ratio of 7∶3. All patients were followed up for 1 year via outpatient visits or telephone interviews, and were categorized into a recurrence group and a non-recurrence group based on their disease status. Clinical data and serum levels of thymus and activation-regulated chemokine (TARC), C-C motif chemokine ligand 19 (CCL19), and leucine-rich α-2-glycoprotein 1 (LRG1) were compared between the two groups. Logistic regression analysis was performed to identify the influencing factors of PV recurrence. Four machine learning models were constructed, and receiver operating characteristic (ROC) curves and area under the curve (AUC) were used to evaluate their predictive performance. Results: Among the 258 PV patients, 8 were lost to follow-up, and 120 experienced recurrence. In both the training set and test set, the recurrence group showed significantly higher serum levels of TARC, CCL19, and LRG1, as well as higher Self-Rating Anxiety Scale (SAS) scores, Self-Rating Depression Scale (SDS) scores, and incidence of sleep disorders compared with the non-recurrence group (Ps＜0.05). Logistic regression analysis confirmed four variables—serum TARC, CCL19, LRG1 levels, and sleep disorders—as independent influencing factors of PV recurrence, which were used for model construction. ROC curve analysis revealed that the random forest (RF) algorithm model achieved higher AUC values in both the training set and test set than the deep learning (DL), gradient boosting machine (GBM), and generalized linear model (GLM) (Ps＜0.05). Conclusion: The random forest algorithm model established based on serum TARC, CCL19, LRG1 levels and sleep disorders can effectively predict the recurrence risk of PV.

Key words: psoriasis vulgaris, recurrence, early warning model, TARC, CCL19, LRG1