利奈唑胺致老年患者贫血的危险因素分析及风险预测模型构建

Risk factors of linezolid-induced anemia and the establishment of the risk prediction model in elderly patients

目的探讨老年患者应用利奈唑胺发生相关贫血的危险因素,并建立贫血的风险预测模型。方法回顾性收集2014年1月至2020年8月于复旦大学附属中山医院接受利奈唑胺治疗的115例老年患者的临床资料,包括人口统计学特征、实验室检查指标、合并疾病和联合用药等情况。根据发生贫血与否,将患者分为贫血组(n=21)与非贫血组(n=94)。比较2组间相关指标的差异,通过多因素logistic回归分析筛选老年患者利奈唑胺相关贫血的独立危险因素,并建立预测模型,采用ROC曲线评估模型的预测效能。结果共115例老年患者,其中男性75例(65.2%),21例发生贫血(18.3%)。单因素及logistic多因素回归分析表明,用药天数、估算的肾小球滤过率、联合使用胺碘酮及联合使用他克莫司/西罗莫司是老年患者发生利奈唑胺相关贫血的独立影响因素(P<0.05)。Logistic回归方程为Logit(P)=0.160XT+2.580XAm+2.807XTac-0.024XeGFR-1.994,联合预测因子的ROC曲线下面积为0.843,灵敏度为90.5%,特异度为70.5%;当约登指数最大(0.607)时,对应的ROC曲线上的临界值为0.5。结论老年患者中,利奈唑胺相关贫血的独立影响因素为用药时间、估算的肾小球滤过率、联用胺碘酮及他克莫司/西罗莫司;预测模型计算结果实际值大于临界值时,表明老年患者发生利奈唑胺相关贫血的风险较高,应加强血红蛋白的监测。

Objective To explore the risk factors for linezolid-induced anemia and establish the risk prediction model in elderly patients.Methods The data of elderly patients treated with linezolid from January 2014 to August 2020 were retrospectively collected including demographic characteristic,baseline laboratory results,combind diseases,and concomitant medications.According to the occurrence of anemia,the patients were divided into anemia group and non-anemia group.The differences of related indexes between the two groups were compared.The logistic regression analysis was used to screened the independent risk factors for linezolid-related anemia in elderly patients.Then a risk prediction model was established,and the ROC curve was used to evaluate the predictive value of the model.Results A total of 115 elderly patients were included,among whom,75 were males(65.2%)and 21(18.3%)developed anemia during the linezolid treatment.The univariate and multivariate logistic regression analyses showed that the duration of linezolid therapy,estimated glomerular filtration rate(eGFR),concomitantwith amiodarone,and concomitant with tacrolimus/sirolimus were independent influence factors for linezolid-related anemia.Logistic regression formula was further constructed according to the above influence factors:Logit(P)=0.160XT+2.580XAm+2.807XTac-0.024XeGFR-1.994.The area under the ROC curve of the joint predictor was 0.843,with a certain predictive power,and the sensitivity and specificity of it were respectively 90.5%and 70.5%.When the Youden index is the largest(0.607),the corresponding cut-off value on the ROC curve was 0.5.Conclusion The independent influence factors for linezolid-related anemia in elderly patients were duration of linezolid therapy,eGFR,concomitant with amiodarone,and concomitant with tacrolimus/sirolimus.It would indicate that the elderly patients have a higher risk of linezolid-associated anemia when the actual value of the formula is greater than the cut-off value,and the hemoglobin should be monitored.