文献速递治疗药物监测与药物基因组学特殊人群用药

机制引导机器学习用于肾移植早期他克莫司个体化剂量调整

Mechanism-informed machine learning for individualized tacrolimus dose adjustment in the early post-kidney transplant period.
PubMed 文献 · Br J Clin Pharmacol · 2026-01-13

本研究基于1624例肾移植受者术后7天内的4311个他克莫司谷浓度,开发了融合药代动力学原理的GRU和XGBoost机器学习模型以预测谷浓度并推荐剂量。机制引导GRU模型在内外部验证中均优于贝叶斯群体药代动力学模型,且当临床剂量落在GRU推荐范围内时,后续谷浓度达标率显著提高。

要点

  • 机制引导GRU模型预测他克莫司谷浓度优于贝叶斯PopPK模型。
  • 临床剂量落在GRU推荐范围时,后续谷浓度达标率51.3% vs 总体37.0%。
  • 融合药代动力学原理未降低模型准确性且提升可解释性。

核心结论

  • 机制引导机器学习模型可为肾移植早期他克莫司个体化给药提供稳健且可解释的方法。
  • 该模型有潜力通过精准剂量调整提高治疗目标达标率。

研究设计与样本

  • 研究纳入1624例肾移植受者术后7天内的4311个他克莫司谷浓度。
  • 开发了融合线性药代动力学原理的GRU和XGBoost两种机器学习模型。
  • 通过内部交叉验证和外部验证比较机制引导模型、贝叶斯PopPK模型与纯数据驱动ML模型。

主要发现

  • 机制引导GRU模型内部验证MSE=7.81 ng2/mL2、R2=0.537,优于贝叶斯PopPK的MSE=9.27 ng2/mL2、R2=0.450。
  • 外部验证GRU模型MSE=6.09 ng2/mL2、R2=0.397,优于贝叶斯PopPK的MSE=8.96 ng2/mL2、R2=0.211。
  • 机制引导XGBoost模型表现与GRU模型相当。
  • 临床常规TDM确定的剂量落在GRU推荐范围内时,后续谷浓度达8-12 ng/mL目标的比例为51.3%,总体为37.0%(p<0.01)。

关键数值与推荐

  • 他克莫司治疗目标谷浓度为8-12 ng/mL。
  • GRU模型推荐剂量范围内给药时后续谷浓度达标率51.3%,总体37.0%(p<0.01)。

临床怎么用

  • 在肾移植术后早期可借助机制引导ML模型预测他克莫司谷浓度并辅助剂量推荐。
  • 将临床给药剂量与模型推荐范围比对,有助于提高谷浓度达标率。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

AIM: Tacrolimus dosing in the early post-kidney transplant period is challenging due to a narrow therapeutic index and substantial interindividual pharmacokinetic (PK) variability. This study aimed to develop and validate mechanism-informed machine learning (ML) models to support individualized tacrolimus dosing during this critical period. METHODS: A total of 4311 tacrolimus trough concentrations (Ctrough) within 7 days post-transplant were obtained from 1624 kidney transplant recipients. Two ML models, Gated Recurrent Unit (GRU) and eXtreme Gradient Boosting (XGBoost), were developed to predict Ctrough and recommend doses to achieve target levels. Both models incorporated PK principles based on linear pharmacokinetics. Model performance was compared to a traditional Bayesian population PK (PopPK) model and purely data-driven ML models via internal cross-validation and external validation. RESULTS: The mechanism-informed GRU model outperformed the Bayesian PopPK model in both internal validation (MSE = 7.81 vs. 9.27 ng2/mL2, R2 = 0.537 vs. 0.450) and external validation (MSE = 6.09 vs. 8.96 ng2/mL2, R2 = 0.397 vs. 0.211). The mechanism-informed XGBoost model performed comparably to the GRU model. The incorporation of PK principles enhanced model interpretability and generalizability without reducing accuracy. When clinically administered doses, determined by conventional therapeutic drug monitoring, fell within the GRU model's recommended range, subsequent Ctrough reached the therapeutic target (8-12 ng/mL) in 51.3% of cases, compared to 37.0% overall (p < 0.01). CONCLUSION: Mechanism-informed ML models offer a robust and interpretable approach for individualized tacrolimus dosing, with the potential to improve therapeutic target attainment by enabling accurate dose adjustments in the early post-transplant period.

来源:Br J Clin Pharmacol,查看原文。版权归原作者所有。

引用

  1. Yu H, Qin Z, Smith LS, et al. Mechanism-informed machine learning for individualized tacrolimus dose adjustment in the early post-kidney transplant period. Br J Clin Pharmacol. 2026. PMID: 41531243. doi:10.1002/bcp.70448

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。