药源性血小板减少症的危险因素分析与风险预测模型的构建

Analysis of risk factors and construction of a risk prediction model for drug-induced thrombocytopenia

目的 了解药源性血小板减少症(DITP)的临床特征、危险因素,并建立DITP风险预测模型。方法 通过临床药物不良事件主动监测与智能评估警示系统-Ⅱ回顾性分析2021年1—12月解放军总医院住院患者DITP的发生情况。应用倾向性评分匹配(PSM),为每例阳性病例匹配3例对照病例,通过单因素分析联合最大似然比检验筛选预测变量,构建DITP列线图预测模型。结果 共监测住院患者88 151例,经判定DITP阳性病例478例(0.5%);涉及主要怀疑药品87种,主要为抗肿瘤药(57.3%,274/478)、抗菌药物(25.1%,120/478)和抗血栓药(12.6%,60/478)。通过PSM后,DITP组382例,N-DITP组1146例患者。多因素logistic回归结果显示,住院时间≥14 d、联合使用抗菌药物、基线γ-谷氨酰转移酶(GGT)≥50 U/L是DITP的危险因素,基线血小板计数(PLT)≥150×109/L是DITP的保护因素(P<0.05)。将以上4个变量用于构建列线图模型,模型的AUC为0.821(95%CI:0.799~0.844),校准曲线显示列线图模型校准度良好(P>0.05)。结论 导致住院患者DITP的主要药物为抗肿瘤药、抗菌药物和抗血栓药;基于住院时间、联合使用抗菌药物、基线GGT和PLT构建的列线图模型,有助于预测DITP风险,为DITP防治提供参考。

Objective To investigate the clinical characteristics and risk factors of drug-induced thrombocytopenia(DITP)and establish a risk prediction model for DITP.Methods The occurrence of DITP in hospitalized patients from January to December 2021 in the Chinese PLA General Hospital was retrospectively analyzed using the adverse drug events active surveillance and assessment system-Ⅱ.Propensity score matching(PSM)was applied to match 3 control cases for each positive case.Predictor variables were screened through univariate analysis combined with the maximum likelihood ratio test,and a nomogram prediction model for DITP was constructed.Results A total of 88151 hospitalized patients were monitored,and 478(0.5%)were diagnosed as DITP-positive cases.A total of 87 suspected drugs were involved,mainly including antineoplastic drugs(57.3%,274/478),antimicrobials(25.1%,120/478),and antithrombotic drugs(12.6%,60/478).After PSM,382 patients were in the DITP group and 1146 patients were in the N-DITP group.Multivariate logistic regression analysis showed that hospital stay≥14 days,concomitant use of antimicrobials,and baselineγ-glutamyltransferase(GGT)≥50 U/L were risk factors for DITP,while baseline platelet count(PLT)≥150×109/L was a protective factor(P<0.05).These four variables were used to construct a nomogram model,with the AUC of 0.821(95%CI:0.799 to 0.844),and the calibration curve indicated good calibration of the nomogram model(P>0.05).Conclusion The main drugs causing DITP in hospitalized patients are antineoplastic drugs,antimicrobials,and antithrombotic drugs.The nomogram model based on hospital stay,concomitant use of antimicrobials,baseline GGT and PLT can help predict the risk of DITP,providing a reference for the prevention and treatment of DITP.