更多内容请点击

中国全科医学

• •

肺结核患者氟喹诺酮类耐药影响因素预测模型的构建与验证：基于LASSO-Logistic回归模型

秦娅莉1，陈静1，2*，李军1，王明栋3，欧维正3，邱继瑶2，彭燕清2   

  1. 1.561113 贵州省贵阳市，贵州医科大学公共卫生与健康学院 环境污染与疾病监控教育部重点实验室；2.550003 贵州省贵阳市公共卫生救治中心结核科；3.550003 贵州省贵阳市公共卫生救治中心检验科
  • 收稿日期:2023-12-28 修回日期:2024-02-07 接受日期:2024-02-28
  • 通讯作者: 陈静
  • 基金资助:
    “十三五”国家重大新药创制项目（2017ZX09304009）；贵阳市科技计划项目（筑科合同［2018］1-47 号）

Construction and Validation of a Predictive Model of Influencing Factors for Fluoroquinolone Resistance in Patients with Pulmonary Tuberculosis：Based on the LASSO-Logistic Regression Model

QIN Yali1，CHEN Jing1，2*，LI Jun1，WANG Mingdong3，OU Weizheng3，QIU Jiyao2，PENG Yanqing2   

  1. 1.School of Public Health，the Key Laboratory of Environmental Pollution Monitoring and Disease Control，Ministry of Education，Guizhou Medical University，Guiyang 561113，China；2.Department of Tuberculosis，Guiyang Public Health Clinical Center，Guiyang 550003，China；3.Department of Laboratory，Guiyang Public Health Clinical Center，Guiyang 550003，China
  • Received:2023-12-28 Revised:2024-02-07 Accepted:2024-02-28
  • Contact: CHEN Jing

PDF
分享到

摘要： 背景　利福平耐药/耐多药结核病（RR/MDR-TB）治疗困难，治愈率低，且传染性强，氟喹诺酮类（FQs）作为治疗 RR/MDR-TB 的核心药物，耐药趋势严峻，对 FQs 影响因素分析有助于提高 RR/MDR-TB 的治愈率，并控制准广泛耐药（pre-XDR）和广泛耐药结核病（XDR）的发生。目的　分析住院肺结核患者 FQs 耐药情况及影响因素，构建 FQs 耐药危险因素的列线图（Nomogram）预测模型并进行验证。方法　回顾性选取于 2021 年 1 月—2022 年 2 月在贵阳市公共卫生救治中心住院且有药物敏感试验结果的 583 例肺结核患者为研究对象。根据患者治疗史将 583 例肺结核患者分为初治组（296 例）和复治组（287 例）；根据患者 FQs 耐药情况将 583 例肺结核患者分为 FQs 耐药组（63 例）和 FQs 敏感组（520 例）。分析 583 例患者对 13 种抗结核药物总耐药分布情况，比较 FQs 耐药组与 FQs 敏感组肺结核患者的基线特征。采用 LASSO 回归模型筛选特征变量后，行多因素 Logistic 回归分析 FQs 耐药的独立危险因素，并构建 Nomogram 预测模型；采用受试者工作特征（ROC）曲线下面积（AUC）、校准曲线对其进行验证。结果　583 例患者中 FQs 敏感为 520 例，耐药 63 例，耐药率为 10.81%，仅次于一线抗结核药异烟肼、利福平、链霉素、乙胺丁醇总耐药率（36.36%、32.76%、21.61%、12.86%）。复治组患者利福平、异烟肼、乙胺丁醇、链霉素、左氧氟沙星、莫西沙星、利福平耐药（RR）、耐多药（MDR）、pre-XDR 耐药率高于初治组（P<0.05）。FQs 耐药组患者少数民族、复治、艾滋病、吸毒史、空洞、咯血、不规则抗痨史、MDR 占比高于 FQs 敏感组，（P<0.05）。LASSO回归筛选出 6 个变量：民族、治疗史、艾滋病、吸毒史、咯血、MDR；多因素 logistic 回归分析结果显示，少数民族（OR=2.313，95%CI=1.153~4.640，P=0.018）、复治（OR=1.892，95%CI=1.005~3.560，P=0.048）、咯血（OR=1.941，95%CI=1.087~3.465，P=0.025）、MDR（OR=3.342，95%CI=2.398~7.862，P<0.001） 是 肺 结 核 患 者 FQs 耐 药 的 独 立危险因素；Logistic 回归方程 Logit（P）=-3.571+0.838× 民族 +0.638× 治疗史 +0.663× 咯血 +1.468×MDR，基于此构建风险 Nomogram 预测模型，AUC 为 0.796（95%CI=0.717~0.876），Bootstrap 法验证平均绝对误差为 0.015，通过Hosmer-Lemeshow 拟合优度检验，预测模型有较好的校准能力（χ2=3.426，P=0.489>0.05）。结论　肺结核患者 FQs耐药率较高，少数民族、复治、咯血、MDR 是肺结核患者 FQs 耐药的独立危险因素，构建 Nomogram 预测模型对于肺结核患者 FQs 耐药具有较好的预测价值，能够为临床诊断耐药结核病及为 RR/MDR-TB 制订合理治疗方案提供新思路。

关键词: 结核, 肺；氟喹诺酮类；结核分枝杆菌；药物敏感试验；广泛耐药结核；耐多药结核；危险因素；列线图

Abstract: Background　Rifampicin-resistant/multidrug-resistant tuberculosis（RR/MDR-TB）is featured by challenges in the treatment，low cure rate，and high infectivity. Fluoroquinolones（FQs），as the core drugs for the treatment of RR/MDR-TB，have a severe trend of resistance. Analyzing influencing factors for FQs can help to increase the cure rate of RR/MDR-TB and to control the occurrence of the pre-extensive drug resistance（pre-XDR）and extensive drug resistance（XDR）. Objective　To analyze the drug resistance of FQs in hospitalized patients with pulmonary tuberculosis and the influencing factors，and to construct and validate a Nomogram prediction model for the risk factors of drug resistance of FQs. MethodsA total of 583 patients with pulmonary tuberculosis who were hospitalized in Guiyang Public Health Clinical Center from January 2021 to February 2022 and tested for drug sensitivity were retrospectively selected as study subjects. They were divided into the initial treatment group（296 patients）and the retreatment group（287 patients）according to the history of previous treatment. Moreover，they were divided into the FQs-resistant group（63 patients）and FQs-sensitive group（520 patients）according to their FQs-resistance status. The distribution of total resistance to 13 antituberculosis drugs in 583 patients was analyzed，and the baseline characteristics of patients in the FQs-resistant group and FQs-sensitive group were compared. After screening the characteristic variables using least absolute shrinkage and selection operator（LASSO）regression model，multivariate Logistic regression was performed to analyze the independent risk factors for the resistance of FQs. A Nomogram prediction model was constructed，and its performance was validated by calculating the area under the curve（AUC）of receiver operating characteristic（ROC），and plotting the calibration curve. Results　Among 583 patients，520 cases were sensitive to FQs and 63 cases were resistant（resistance rate of 10.81%）. The resistance rate of FQs was secondary to the total resistance rate of first-line antituberculosis drugs，including the isoniazid（36.36%），rifampicin（32.76%），streptomycin（21.61%），and ethambutol（12.86%）. The resistance rates of rifampicin，isoniazid，ethambutol，streptomycin，levofloxacin，moxifloxacin and rifampicin resistance（RR），multidrug resistance（MDR），and pre-XDR were significantly higher in patients of the retreatment group than those of the initial treatment group（P<0.05）. The proportions of patients with ethnic minorities，retreatment，acquired immunodeficiency syndrome（AIDS），history of drug abuse，cavitation，hemoptysis，history of irregular anti-TB and MDR were significantly higher in the FQs-resistance group than those of the FQs-sensitive group（P<0.05）. Through LASSO regression，six variables of ethnicity，treatment history，AIDS，drug abuse history，hemoptysis，and MDR were screened out as influencing factors. Multivariate Logistic regression analysis showed that ethnicity（OR=2.313，95% CI=1.153-4.640，P=0.018），retreatment（OR=1.892，95%CI=1.005-3.560，P=0.048），hemoptysis（OR=1.941，95%CI=1.087-3.465，P=0.025），and MDR（OR=3.342，95% CI=2.398-7.862，P<0.001）were the independent risk factors for FQs resistance in patients with pulmonary tuberculosis. Logistic regression equation Logit（P）=-3.571+0.838×ethnicity+0.638×treatment history+0.663×hemoptysis+1.468×MDR. Based on which a risk Nomogram prediction model was constructed with an AUC of 0.796（95%CI=0.717-0.876）. The Bootstrap method validated the mean absolute error of 0.015，and the predictive model had good calibration ability by the Hosmer-Lemeshow goodness-of-fit test（χ2=3.426，P=0.489）. Conclusion　Our findings suggest a high resistant rate of FQs in patients with pulmonary tuberculosis. Ethnic minorities，retreatment，hemoptysis，and MDR are independent risk factors for FQs resistance in patients. The constructed Nomogram prediction model has a good predictive value for FQs resistance in patients with pulmonary tuberculosis. Our study offers new insights into the clinical diagnosis of drug-resistant tuberculosis and the development of rational treatment regimens for RR/MDR-TB.

Key words: Tuberculosis, pulmonary；Fluoroquinolones；Mycobacterium tuberculosis；Drug susceptibility testing；Extensively drug-resistant tuberculosis；Multidrug-resistant tuberculosis；Risk factors；Nomograms

中图分类号: 

冀ICP备12018774号-6　冀公网安备 13040302001295号　网站版权 © 《中国全科医学》编辑部　技术支持：北京玛格泰克科技发展有限公司