利奈唑胺相关血小板减少风险预测模型构建及应用

Construction and Application of a Risk Prediction Model for Linezolid-Related Thrombocytopenia

目的建立利奈唑胺相关血小板减少的联合因素风险预测模型,根据联合因素风险阈值预测初次治疗药物监测(TDM)的时间。方法选取新疆某三甲医院2017年1月至2022年1月使用利奈唑胺治疗感染性疾病的患者149例,根据是否发生血小板减少分为血小板减少组(61例)和非血小板减少组(88例),采用Logistic回归分析法筛选利奈唑胺相关血小板减少的危险因素,建立联合因素的风险预测模型。根据联合因素风险阈值预测初次TDM时间,并进行验证。结果61例(40.94%)患者发生了血小板减少症,疗程不低于12 d[OR=6.620,95%CI(2.338,18.74),P<0.001],估算肾小球滤过率不超过60 mL/(min·1.73 m^(2))[OR=7.352,95%CI(1.723,31.37),P=0.007],内生肌酐清除率(CCr)不超过60 mL/min[OR=4.464,95%CI(1.349,14.77),P=0.014]为利奈唑胺相关血小板减少的独立危险因素。联合因素(疗程和CCr)的受试者工作特征(ROC)曲线下面积(AUC)为0.9042,95%CI为(0.8566,0.9519),敏感度为86.89%,特异度为80.68%,临界值为4.871;确定联合因素的风险阈值为5.0。结论所建立的联合因素风险预测模型具有较好的诊断效能,可用于反向预测初次TDM的时间,提高利奈唑胺临床应用的有效性和安全性。

Objective To establish a combined factor risk prediction model for linezolid-related thrombocytopenia,and to predict the time of first therapeutic drug monitoring(TDM)according to the combined factor risk threshold.Methods A total of 149 patients with infectious diseases treated with linezolid from January 2017 to January 2022 in a Grade-A hospital in Xinjiang were selected and divided into the thrombocytopenia group(n=61)and the non-thrombocytopenia group(n=88)according to whether they had thrombocytopenia or not.The risk factors associated with linezolidin-related thrombocytopenia were screened by the Logistic regression analysis,and a combined factor risk prediction model was established.The time of the first TDM was predicted and validated according to the the combined factor risk threshold.Results A total of 61 patients(40.94%)developed thrombocytopenia,and the course of treatment≥12 d[OR=6.620,95%CI(2.338,18.74),P<0.001],epidermal growth factor receptor(eGFR)≤60 mL/(min·1.73 m^(2))[OR=7.352,95%CI(1.723,31.37),P=0.007],and creatinine clearance rate(CCr)≤60 mL/min[OR=4.464,95%CI(1.349,14.77),P=0.014]were independent risk factors for linezolidin-related thrombocytopenia.The area under the receiver operating characteristic(ROC)curve(AUC)of the combined factors(course of treatment and CCr)was 0.9042,with a 95%CI of 0.8566-0.9519,a sensitivity of 86.89%,a specificity of 80.68%,and a critical value of 4.871.The risk threshold for determining the combined factors was 5.0.Conclusion The combined factor risk prediction model has good diagnostic efficacy,which can be used to reverse predict the time of the first TDM,and improve the effectiveness and safety of the clinical application of linezolid.