摘要
目的 构建基于血清和关节液炎性标志物诊断关节置换术后假体周围感染(prosthetic joint infections,PJI)的列线图诊断模型,并验证其预测能力.方法 回顾性收集2015年1月至2020年6月于重庆医科大学附属第一医院骨科诊断为PJI或无菌性松动的181例患者临床资料作为建模组.通过lasso回归、单因素及多因素分析筛选出诊断PJI的最佳指标.通过综合考量各项指标的权重及内在联系,构建列线图诊断模型,并以此开发临床决策支持系统(clinical decision support system,CDSS).前瞻性收集2020年7月至2022年12月于重庆医科大学附属第一医院骨科诊断为PJI或无菌性松动的49例患者作为验证组,通过受试者工作特征曲线(receiver operating characteristic curve,ROC)等方法对列线图模型的诊断性能进行外部验证.结果 181例建模组中PJI患者85例,49例验证组中PJI患者23例.在27个潜在因素中通过las-so 回归分析发现体质指数(body mass index,BMI)、血液指标[包括血小板、淋巴细胞绝对值、干扰素γ(interferon γ,IFN-γ)、红细胞沉降率、IL-6、C反应蛋白、D-二聚体]及关节液指标(包括C反应蛋白、IL-1、IL-4、IL-6、多核中性粒细胞百分比、CD64)可能是诊断PJI的潜在指标.单因素回归分析发现血液学指标(包括红细胞沉降率、C反应蛋白、IL-6、D-二聚体)和关节液指标(包括C反应蛋白、关节液CD64指数、C反应蛋白、IL-1、IL-4、IL-6、关节液多形核白细胞百分比)的差异有统计学意义(P＜0.05).多因素回归分析筛选出血清IL-6、D-二聚体、关节液CD64指数、C反应蛋白、IL-1、IL-4、IL-6及关节液多形核白细胞百分比等指标,以此为基础构建列线图模型和CDSS系统.验证组外部验证ROC曲线下面积(area under the curve,AUC)为0.978,内部验证AUC为0.995;校准曲线C指数为99.50％,内部验证C指数为99.53％,提示列线图模型在临床应用中具有良好的预测能力.结论 基于多个诊断指标诊断PJI的列线图显示出良好的诊断性能.通过列线图构建的CDSS系统将有助于临床医生更好地诊断PJI,及时做出合理决策.
Abstract
Objective To construct a column-line diagram diagnostic model based on serum and joint fluid inflammatory markers for the diagnosis of periprosthetic joint infections(PJI)after joint arthroplasty and to validate its predictive ability.Meth-ods The clinical data of 181 patients diagnosed with PJI or aseptic loosening in the Department of Orthopedics of the First Affili-ated Hospital of Chongqing Medical University from January 2015 to June 2020 were retrospectively collected as a modeling group.The best indicators for diagnosing PJI were screened by lasso regression,single-factor and multifactor analysis.By compre-hensively considering the weights and intrinsic connections of the indicators,a column-line diagram diagnostic model was con-structed and used to develop a clinical decision support system(CDSS).Prospectively,the clinical data of patients diagnosed with PJI or aseptic loosening in the Department of Orthopedics of the First Hospital of Chongqing Medical University from July 2020 to December 2022 were collected as a validation group,and the diagnostic performance of the column-line diagram model was exter-nally validated by methods such as receiver operating characteristic curve(ROC).Results There were 85 cases of PJI in the 181 cases modeling group and 23 cases of PJI in the 49 cases validation group.Among the 27 potential factors analyzed by lasso regres-sion analysis,body mass index(BMI),blood tests including platelet(PLT),absolute lymphocyte value,interferon γ(IFN-γ),ESR,IL-6,C-reactive protein,D-dimer,and joint fluid tests including C-reactive protein,IL-1,IL-4,IL-6,percentage of multinucleated neutrophils(PMN％),and CD64 may be potential indicators for the diagnosis of PJI.Univariate found significant differences be-tween hematologic tests including sedimentation,C-reactive protein,IL-6,D-dimer and joint fluid tests including C-reactive pro-tein,joint fluid CD64 index,C-reactive protein,IL-1,IL-4,IL-6,PMN％(P＜0.05).Further multifactorial regression analysis screened serum 1L-6,D-dimer,joint fluid CD64 index,C-reactive protein,IL-1,IL-4,IL-6,and percentage of multinucleated neu-trophils,and based on that,the column-line graph model and CDSS system were constructed.The area under the ROC in the vali-dation group was 0.978,and the AUC in the internal validation was 0.995;the C-index of the calibration curve was 99.50％,and the C-index of the internal validation was 99.53％,suggesting that the column-line diagram model has a good predictive ability.Conclusions The column-line diagram for diagnosing PJI based on multiple diagnostic indicators showed good diagnostic perfor-mance.The CDSS system constructed by column-line diagrams could assist clinicians in diagnosing PJI and making reasonable strategies in time.
NETL
NSTL
万方数据