
浏览全部资源
扫码关注微信
1.中国科学技术大学附属第一医院(安徽省立医院);药学部,合肥 230001
2.精准药物制剂与临床药学安徽省重点实验室,合肥 230001
副主任药师,博士。研究方向:临床药学。E-mail:czl0808@ustc.edu.cn
主任药师,博士。研究方向:临床药学。E-mail:tangliqin@ustc.edu.cn
收稿:2026-04-10,
修回:2026-06-17,
录用:2026-07-10,
网络首发:2026-07-24,
纸质出版:2026-07-30
移动端阅览
陈昭琳,朱鹏里,倪伟建,等.基于群体药动学信息与机器学习的他克莫司血药浓度预测模型研究 [J].中国药房,2026,37(14):1898-1904.
CHEN Zhaolin,ZHU Pengli,NI Weijian,et al.Prediction model of tacrolimus blood concentration based on population pharmacokinetics information and machine learning[J].ZHONGGUO YAOFANG,2026,37(14):1898-1904.
陈昭琳,朱鹏里,倪伟建,等.基于群体药动学信息与机器学习的他克莫司血药浓度预测模型研究 [J].中国药房,2026,37(14):1898-1904. DOI: 10.6039/j.issn.1001-0408.2026.14.17.
CHEN Zhaolin,ZHU Pengli,NI Weijian,et al.Prediction model of tacrolimus blood concentration based on population pharmacokinetics information and machine learning[J].ZHONGGUO YAOFANG,2026,37(14):1898-1904. DOI: 10.6039/j.issn.1001-0408.2026.14.17.
目的
2
构建基于群体药动学(PPK)信息与机器学习的他克莫司谷浓度预测模型,评估其在临床治疗药物监测(TDM)数据中的预测性能,为肾移植术后他克莫司个体化剂量调整提供循证依据。
方法
2
采用回顾性队列研究设计,纳入2025年1-6月在中国科学技术大学附属第一医院(安徽省立医院)治疗的肾移植受者的连续TDM记录,构建最终分析集。围绕前次TDM浓度、前次浓度/剂量(
C
/
D
)比、个体预测暴露量(IPRE)等29个输入变量,系统比较13种常用回归算法模型及1种加权集成模型。以患者为单位,按80∶20将数据集分为训练集与测试集,在训练集上开展基于患者分组的5折交叉验证,最终在获得的
评估点上采用决定系数(
R
2
)、平均绝对误差(MAE)、均方根误差(RMSE)、Pearson相关系数(Corr)及±20%/±30%准确率评价模型性能。采用沙普利加性解释(SHAP)法开展模型可解释性分析,明确核心预测特征。
结果
2
共纳入83例肾移植受者的809条连续TDM记录,其中训练集66例、测试集17例。SVR模型表现最优,
R
2
为0.649 8,MAE为1.126 0 ng/mL,RMSE为1.478 6 ng/mL,Corr为0.819 1,±20%准确率为59.68%,±30%准确率为79.03%。Weighted Ensemble模型与Extra Trees模型位列其后。SHAP可解释性分析显示,贡献排前6名的特征依次为IPRE、前次TDM浓度、前次
C
/
D
比、剂量标准化浓度、红细胞压积(HCT)、浓度-剂量系数(
θ
cd
),其中IPRE为核心预测特征(平均SHAP值0.915 4)。
结论
2
融合PPK信息的机器学习模型在当前TDM队列中展现了较好的他克莫司浓度预测能力;SVR为最优模型,IPRE为模型核心预测特征,少数核心PK派生特征(IPRE、剂量标准化浓度、
θ
cd
)与临床特征(前次TDM浓度、HCT)共同构成模型预测的主要信息源。
OBJECTIVE
2
To construct a prediction model for tacrolimus trough concentration based on population pharmacokinetics (PPK) information and machine learning, evaluate its predictive performance in clinical therapeutic drug monitoring (TDM) data, and provide evidence-based support for individualized dose adjustment of tacrolimus after renal transplantation.
METHODS
2
A retrospective cohort study design was adopted, and consecutive TDM records of kidney transplant recipients treated in the First Affiliated Hospital of University of Science and Technology of China (Anhui Provincial Hospital) from January to June 2025 were included to construct the final analysis set. Based on 29 input variables including previous TDM concentration, previous concentration/dose (
C
/
D
) ratio, and individual predicted exposure (IPRE), 13 commonly used regression algorithm models and 1 weighted ensemble model were systematically compared. The dataset was split into training set and test set in an 80∶20 ratio by patient. Five-fold cross-validation based on patient grouping was performed on the training set, and the model performance was evaluated
on the obtained evaluation points using the coefficient of determination (
R
2
), mean absolute error (MAE), root mean square error (RMSE), Pearson correlation coefficient (Corr) and ±20%/±30% accuracy. The SHapley Additive exPlanations (SHAP) method was used for model interpretability analysis to identify core predictive features.
RESULTS
2
A total of 809 consecutive TDM records from 83 kidney transplant recipients were included, with 66 cases in the training set and 17 cases in the test set. The SVR model showed the best performance, with
R
2
of 0.649 8, MAE of 1.126 0 ng/mL, RMSE of 1.478 6 ng/mL, Corr of 0.819 1, ±20% accuracy of 59.68%, and ±30% accuracy of 79.03%. The Weighted Ensemble model and Extra Trees model ranked second and third, respectively. SHAP interpretability analysis showed that the top 6 features contributing to the prediction were IPRE, previous TDM concentration, previous
C/D
ratio, dose-normalized concentration, hematocrit (HCT), and concentration-dose coefficient (
θ
cd
), among which IPRE was the core predictive feature (SHAP mean value 0.915 4).
CONCLUSIONS
2
The machine learning model incorporating PPK information demonstrates good predictive ability for tacrolimus concentration in the current TDM cohort. SVR is the optimal model, IPRE is the core predictive feature of the model, and a few core PK-derived features (IPRE, dose-normalized concentration,
θ
cd
) together with clinical features (previous TDM concentration, HCT, etc.) constitute the main information sources for model prediction.
Dieter R , Jaffe I S , Mattoo A . Tacrolimus in solid organ transplantation:historical impact,current use,and future directions in immunosuppression [J ] . J Pharm Pract , 2026 : 8971900261427132 .
陈冰 , 左笑丛 , 李新刚 , 等 . 模型引导的抗排异治疗患者他克莫司精准用药专家共识 [J ] . 中国临床药理学与治疗学 , 2025 , 30 ( 4 ): 433 - 445 .
Cheng X , Chen Y H , Zhang L , et al . Influence of CYP3A5,IL-10 polymorphisms and metabolism rate on tacrolimus exposure in renal post-transplant recipients [J ] . Pharmacogenomics , 2022 , 23 ( 18 ): 961 - 972 .
Hirai T , Morikawa Y , Onishi R , et al . Impact of glycaemic control and CYP3A5 polymorphisms on tacrolimus trough concentrations after adult kidney transplantation [J ] . Br J Clin Pharmacol , 2023 , 89 ( 6 ): 1852 - 1861 .
Chen H S , Liu S , Yu L L , et al . Association between tacrolimus intrapatient variability and clinical outcomes in kidney transplantation:a retrospective cohort study [J ] . Transl Androl Urol , 2025 , 14 ( 10 ): 3203 - 3211 .
Tsyplakova A , Catic-Djorđevic A , Stefanović N , et al . Optimizing mycophenolate therapy in renal transplant patients using machine learning and population pharmacokinetic modeling [J ] . Med Sci , 2025 , 13 ( 4 ): 235 .
陈惠敏 , 汪洋 , 高柳柳 , 等 . 群体药动学模型引导的机器学习用于预测抽动障碍儿童阿立哌唑清除率 [J ] . 中国医院药学杂志 , 2024 , 44 ( 19 ): 2215 - 2222 .
Huang Q B , Lin X B , Wang Y , et al . Tacrolimus pharmacokinetics in pediatric nephrotic syndrome:a combination of population pharmacokinetic modelling and machine learning approaches to improve individual prediction [J ] . Front Pharmacol , 2022 , 13 : 942129 .
Hoffert Y , Dia N , Vanuytsel T , et al . Model-informed precision dosing of tacrolimus:a systematic review of population pharmacokinetic models and a benchmark study of software tools [J ] . Clin Pharmacokinet , 2024 , 63 ( 10 ): 1407 - 1421 .
Francke M I , Sassen S D T , Lloberas N , et al . A population pharmacokinetic model and dosing algorithm to guide the tacrolimus starting and follow-up dose in living and deceased donor kidney transplant recipients [J ] . Clin Pharmacokinet , 2025 , 64 ( 9 ): 1379 - 1394 .
Lloberas N , Fernández-Alarcón B , Vidal-Alabró A , et al . State of art of dose individualization to support tacrolimus drug monitoring:what’s next? [J ] . Transpl Int , 2025 , 38 : 14201 .
Wang Y P , Lu X L , Shao K , et al . Improving prediction of tacrolimus concentration using a combination of population pharmacokinetic modeling and machine learning in Chinese renal transplant recipients [J ] . Front Pharmacol , 2024 , 15 : 1389271 .
Chen Z L , Cheng X , Zhang L W , et al . The impact of IL-10 and CYP3A5 gene polymorphisms on dose-adjusted trough blood tacrolimus concentrations in early post-renal transplant recipients [J ] . Pharmacol Rep , 2021 , 73 ( 5 ): 1418 - 1426 .
闵建亮 , 陈国栋 . 机器学习在肾移植受者他克莫司剂量预测中的应用 [J ] . 器官移植 , 2025 , 16 ( 6 ): 921 - 930 .
Ben-Fredj N , Dridi I , Dridi I , et al . Comparison of machine learning algorithms and Bayesian estimation in predicting tacrolimus concentration in Tunisian kidney transplant patients during the early post-transplant period [J ] . Eur J Drug Metab Pharmacokinet , 2025 , 50 ( 3 ): 243 - 250 .
Park D J , Kim M , Cho H , et al . Machine learning-based algorithm for tacrolimus dose optimization in hospitalized kidney transplant patients [J ] . Diagnostics (Basel) , 2025 , 15 ( 23 ): 2948 .
赵美姗 , 李伯钦 , 朱一辰 , 等 . 机器学习模型预测他克莫司浓度并建议肾移植受者的最佳剂量 [J/OL ] . 实用器官移植电子杂志 , 2025 , 13 ( 6 ): 503 - 506 . https://med.wanfangdata.com.cn/Paper/Detail?id=PeriodicalPaper_syqgyzdzzz202506004&dbid=WF_QK. DOI:10.3969/j.issn. 2095-5332.2025.06.004 https://med.wanfangdata.com.cn/Paper/Detail?id=PeriodicalPaper_syqgyzdzzz202506004&dbid=WF_QK.DOI:10.3969/j.issn.2095-5332.2025.06.004 .
Hu K , Pan J J , Qu W Q , et al . Weight,CYP3A5 genotype,and voriconazole co-administration influence tacrolimus initial dosage in pediatric lung transplantation recipients with low hematocrit based on a simulation model [J ] . Curr Pharm Des , 2024 , 30 ( 34 ): 2736 - 2748 .
王国徽 , 李兴德 , 潘娅 , 等 . 3款个体化给药工具用于肾移植患者术后他克莫司的预测准确性及影响因素分析 [J ] . 中国药房 , 2024 , 35 ( 24 ): 3023 - 3028 .
Sikma M A , Hunault C C , Huitema A D R , et al . Clinical pharmacokinetics and impact of hematocrit on monitoring and dosing of tacrolimus early after heart and lung transplantation [J ] . Clin Pharmacokinet , 2020 , 59 ( 4 ): 403 - 408 .
Lloberas N , Grinyó J M , Colom H , et al . A prospective controlled,randomized clinical trial of kidney transplant recipients developed personalized tacrolimus dosing using model-based Bayesian prediction [J ] . Kidney Int , 2023 , 104 ( 4 ): 840 - 850 .
Liang H , Xuan Q L , Liu C W , et al . Machine learning-assisted tacrolimus dose optimization in childhood-onset systemic lupus erythematosus through population pharmacokinetic modeling [J ] . Comput Biol Med , 2025 ,196(Pt. B): 110782 .
Yang W Y , Xia R F , Zhao P J , et al . Impact of high tacro-limus exposure on graft function and development of a prediction model in kidney transplant recipients:a single-center retrospective cohort study [J ] . Drug Des Devel Ther , 2025 , 19 : 12035 - 12047 .
0
浏览量
0
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621