目的：建立万古霉素群体药动学（PPK）模型，并评价胱抑素C（Cys C）对万古霉素药动学参数的影响。方法：回顾性收集225例使用万古霉素治疗患者的333次治疗药物监测（TDM）数据，以其性别（Sex）、年龄（Age）、体质量（mT）、血肌酐（Scr）、Cys C等为协变量，采用非线性混合效应模型法建立万古霉素PPK模型，并采用自举法（Bootstrap）和正态化预测分布误差法（NPDE）进行模型的内部验证。另收集27例患者的40次TDM数据进行模型的外部验证，以平均预测误差（MPE）和均方根误差（RMSE）考察模型预测的准确度与精密度。以最终模型预测典型患者（年龄65岁，体质量64 kg，Scr水平为66 μmol/L，给药方案为1 000 mg，q12 h）在不同Cys C水平时的稳态血药谷浓度和表观清除率（CL/F），评估Cys C的变化对万古霉素药动学参数的影响。结果：患者的年龄、体质量以及Scr、Cys C水平均可显著影响万古霉素的CL/F。Bootstrap法验证的稳健率为100％；除V/F个体间变异外，其余药动学参数的相对偏差均小于5％，且最终模型参数估算值均落在Bootstrap法参数估算值的95％置信区间内；NPDE检验显示其方差齐性且符合正态分布（P＞0.05）。外部验证得最简模型的MPE为－1.52 μg/mL、RMSE为6.87 μg/mL，最终模型的MPE为－0.32          μg/mL、RMSE为4.27 μg/mL，后者预测的准确度与精密度均显著提高。当典型患者的Cys C为0.3～4.0 mg/L时，最终模型预测的稳态血药谷浓度为5.25～29.97 μg/mL，CL/F为1.45～8.71 L/h。结论：患者的年龄、体质量以及Scr、Cys C水平均可显著影响万古霉素的药动学参数，且Cys C水平的变化可造成患者体内万古霉素血药浓度的改变。本研究建立的PPK模型具有一定的代表性，可用于万古霉素个体化药动学参数的估算。

OBJECTIVE： To establish population pharmacokinetics （PPK） model of vancomycin so as to evaluate the effects of cystatin C （Cys C）on the pharmacokinetics parameters of vancomycin. METHODS： Totally 333 times therapeutic drug monitoring （TDM） were retrospectively collected from 225 patients who received vancomycin. Using sex， age， body weight（mT）， Scr and Cys C as covariates， PPK model was established by using nonlinear mixed effect model method. Bootstrap method and normal prediction distribution error （NPDE） method were adopted for internal validation of model. Forty times of TDM data were collected from other 27 patients for external validation. Predicted accuracy and precision of model were investigated with mean prediction error （MPE） and root mean square error （RMSE）. The effects of Cys C change on pharmacokinetic parameters of vancomycin were evaluated with steady state trough concentration and apparent clearance rate （CL/F） of vancomycin in typical patient （65 year-old， 64 kg， Scr 66 μmol/L， 1 000 mg， q12 h） forecasted with the final model at different levels of Cys C. RESULTS： CL/F of vancomycin was significantly influenced by age， body weight， the levels of Scr and Cys C.  V/F was equal to 82.5 L. The robust rate verified by Bootstrap method was 100％. Except for the interindividual variation of V/F， the relative bias of other pharmacokinetic parameters was less than 5％， and the estimated parameters of the final model were in the 95％ confidence intervals of estimated values of Bootstrap. NPDE results showed that the homogeneity of variance was consistent with normal distribution （P＞0.05）. In external validation， MPE and RMSE of the simplest model were －1.52 μg/mL and 6.87 μg/mL. MPE and RMSE of the final model were －0.32 μg/mL and 4.27 μg/mL， the accuracy and precision were improved significantly in the final model. When Cys C levle of typical patient was 0.3-4.0 mg/L， the steady state trough concentration predicted by final model were 5.25-29.97 μg/mL and CL/F were 1.45-8.71 L/h. CONCLUSIONS： Age， body weight， the levels of Scr and Cys C significantly influence the pharmacokinetic parameters of vancomycin； moreover， the level of Cys C can change blood concentration of vancomycin. Established PPK model is of great predictive performance， which can be used to estimate the individual pharmacokinetics parameters of vancomycin.