Page 85 - 《中国药房》2026年8期
P. 85

·药物与临床·


          丙戊酸钠致神经重症患者高氨血症的危险因素分析及风险预测

          模型构建
                        Δ

                *
                                               #
          徐 婉 ,伍 锦,毛娇娇,马晶晶,费 姚(苏州大学附属第四医院药学部,江苏 苏州 215000)
          中图分类号  R971+.6      文献标志码  A      文章编号  1001-0408(2026)08-1039-06
          DOI  10.6039/j.issn.1001-0408.2026.08.12

          摘  要  目的  探讨丙戊酸钠(VPA)致神经重症患者高氨血症的危险因素,并构建风险预测模型。方法  回顾性收集2022年1月
          至2025年6月入住苏州大学附属第四医院重症医学科的172例使用VPA的神经重症患者的临床资料,根据血氨水平将其分为高
          血氨组(73例)和正常组(99例)。采用单因素分析及LASSO回归筛选预测变量,再通过多因素Logistic回归分析筛选独立因素并
          据此绘制列线图。采用受试者操作特征(ROC)曲线、校准曲线及决策曲线分析(DCA)评估模型预测性能。结果  结合单因素分
                                                    2
          析及LASSO回归共筛选出体重指数(BMI)≥24.0 kg/m 、合用苯二氮 类药物、VPA血药浓度、血红蛋白、血清尿素、VPA日均剂量
          及白蛋白7个预测变量;多因素Logistic回归分析结果显示,合用苯二氮 类药物、BMI≥24.0 kg/m 、VPA血药浓度、白蛋白、血清
                                                                                     2
          尿素(比值比分别为 1.615、1.538、1.623、1.942、0.637,95% 置信区间分别为 1.128~2.359、1.059~2.251、1.112~2.431、1.106~
          3.598、0.402~0.980)均与VPA致神经重症患者高氨血症显著相关(P<0.05)。基于上述变量所建列线图预测模型的评估结果显
          示,测试集和验证集的ROC曲线下面积分别为0.810和0.844,校正曲线与实际曲线的趋势基本一致,应用该模型可提高临床净获
          益率。结论  合用苯二氮 类药物、BMI≥24.0 kg/m 、高VPA血药浓度、高白蛋白是VPA致神经重症患者发生高氨血症的独立危
                                                 2
          险因素,而高血清尿素则是其独立保护因素;基于上述因素构建的风险预测模型的区分度、一致性、临床实用性均较好,可用于预
          测VPA致神经重症患者发生高氨血症的发生风险。
          关键词  丙戊酸钠;高氨血症;风险预测模型;独立危险因素;LASSO回归分析;Logistic回归分析

          Analysis  of  risk  factors  for  sodium  valproate-induced  hyperammonemia  in  neurocritical  patients  and
          construction of risk prediction model
          XU Wan,WU Jin,MAO Jiaojiao,MA Jingjing,FEI Yao(Dept.  of  Pharmacy,  the  Fourth Affiliated  Hospital  of
          Soochow University, Jiangsu Suzhou 215000, China)

          ABSTRACT   OBJECTIVE To investigate the risk factors for sodium valproate (VPA)-induced hyperammonemia in neurocritical

          patients,  and  to  construct  a  risk  prediction  model.  METHODS  Clinical  data  were  retrospectively  collected  from  172  neurocritical
          patients  who  received  VPA  treatment  in  the  Department  of  Critical  Care  Medicine,  the  Fourth  Affiliated  Hospital  of  Soochow
          University from January 2022 to June 2025. Patients were divided into the hyperammonemia group (73 cases) and the normal group
         (99  cases)  based  on  their  blood  ammonia  levels.  Univariate  analysis  and  LASSO  regression  analysis  were  used  to  screen  for
          predictive  variables.  Independent  factors  were  identified  through  multivariate  Logistic  regression  analysis,  and  a  nomogram  was
          constructed  accordingly.  The  performance  of  the  model  was  evaluated  using  receiver  operating  characteristic (ROC)  curve,
          calibration  curve,  and  decision  curve  analysis (DCA).  RESULTS  Combination  of  univariate  analysis  and  LASSO  regression
          analysis screened out seven predictive variables: body mass index (BMI)≥24.0 kg/m , concomitant use of benzodiazepines, VPA
                                                                           2
          blood  concentration,  hemoglobin,  serum  urea,  average  daily  VPA  dose,  and  albumin.  Multivariate  Logistic  regression  analysis
          showed that concomitant use of benzodiazepines, BMI≥24.0 kg/m , VPA blood concentration, albumin and serum urea level (with
                                                            2
          odds  ratios  of  1.615,  1.538,  1.623,  1.942  and  0.637,  respectively;  95%  confidence intervals of  1.128-2.359,  1.059-2.251,  1.112-
                                                             2.431,  1.106-3.598  and  0.402-0.980,  respectively)  were  all
             Δ 基金项目 国家自然科学基金项目(No.32400789)
             * 第一作者 主 管 药 师 。 研 究 方 向 :临 床 药 学 。 E-mail:     significantly  associated  with  VPA-induced  hyperammonemia  in
          1327914862@qq.com                                  neurocritical  patients  (P<0.05).  The  nomogram  prediction
             #  通信作者 主 管 药 师 。 研 究 方 向 :临 床 药 学 。 E-mail:    model  constructed  based  on  these  variables  was  evaluated,
          466709991@qq.com                                   showing  that  the  area  under  the  ROC  curve  was  0.810  for  the


          中国药房  2026年第37卷第8期                                                China Pharmacy  2026 Vol. 37  No. 8    · 1039 ·
   80   81   82   83   84   85   86   87   88   89   90