血清IL-4、IL-5、IL-13和sIgE与过敏性鼻炎短期疾病控制丧失相关：一项真实世界决策树模型

文献信息

中文摘要

背景： 过敏性鼻炎（AR）是一种常见的慢性炎症性疾病，部分患者在常规药物治疗下症状初步改善后6个月内出现疾病控制丧失。在真实世界实践中，早期识别疾病控制丧失风险较高的患者仍具有挑战性。
方法： 我们开展了一项单中心回顾性观察性队列研究，纳入2019年至2023年间接受治疗并随访6个月的310例确诊AR成人患者。主要结局为6个月内疾病控制丧失，定义为症状初步改善后再次出现中重度鼻炎症状且需要经临床医生核实的管理调整。基线临床和免疫学数据提取自电子健康记录。使用Boruta算法筛选候选预测因子，并开发决策树分类器，采用十折分层交叉验证进行内部验证。在排除合并鼻窦炎患者的仅AR亚组中进行了预设的敏感性分析。
结果： 使用五种Th2相关生物标志物（IL-4、IL-5、IL-13、sIgE和嗜酸性粒细胞计数）的模型对6个月疾病控制丧失显示出可接受的内部区分度，平均AUC为0.854，准确度为0.875，敏感度为0.809，特异度为0.900，F1分数为0.780。校准和决策曲线分析提示其具有初步的风险分层价值。在仅AR的敏感性分析中性能保持相当。
结论： 基线血清IL-4、IL-5、IL-13、sIgE和嗜酸性粒细胞计数与接受常规药物治疗的AR成人患者6个月内疾病控制丧失相关。该决策树模型提供了基于生物标志物的风险分层的可解释但初步的呈现。鉴于回顾性单中心设计和缺乏外部验证，这些发现应被视为假设生成性。在临床实施前需要前瞻性多中心研究和独立外部验证。该模型针对常规药物治疗，并非旨在评估过敏原免疫治疗或鼻部手术后的结局。

英文摘要

BACKGROUND: Allergic rhinitis (AR) is a common chronic inflammatory disease, and a proportion of patients experience loss of disease control within 6 months after initial symptom improvement under routine pharmacotherapy. Early identification of patients at higher risk of loss of disease control remains challenging in real-world practice.
METHODS: We performed a single-center retrospective observational cohort study including 310 adults with confirmed AR treated between 2019 and 2023 and followed for 6 months. The primary outcome was loss of disease control within 6 months, defined as the re-emergence of moderate-to-severe rhinitis symptoms after initial improvement requiring clinician-verified management adjustment. Baseline clinical and immunological data were extracted from electronic health records. Candidate predictors were screened using the Boruta algorithm, and a decision-tree classifier was developed and internally validated using tenfold stratified cross-validation. A prespecified sensitivity analysis was conducted in an AR-only subset excluding patients with concomitant rhinosinusitis.
RESULTS: The model using five Th2-related biomarkers (IL-4, IL-5, IL-13, sIgE, and eosinophil count) showed acceptable internal discrimination for 6-month loss of disease control, with a mean AUC of 0.854, accuracy of 0.875, sensitivity of 0.809, specificity of 0.900, and F1 score of 0.780. Calibration and decision curve analyses suggested preliminary value for risk stratification. Performance remained comparable in the AR-only sensitivity analysis.
CONCLUSION: Baseline serum IL-4, IL-5, IL-13, sIgE, and eosinophil count were associated with loss of disease control within 6 months in adults with AR receiving routine pharmacotherapy. The decision-tree model provided an interpretable but preliminary representation of biomarker-based risk stratification. Given the retrospective single-center design and lack of external validation, these findings should be regarded as hypothesis-generating. Prospective multicenter studies and independent external validation are required before clinical implementation. This model pertains to routine pharmacotherapy and was not designed to evaluate outcomes after allergen immunotherapy or nasal surgery.