Application of Machine Learning Models to Predict Tigecycline Exposure in Critically Ill Patients Using Different Limited Sampling Strategies.
6 affiliations
- 1Department of Pharmacy, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, People's Republic of China.
- 2Research Center for Clinical Pharmacy, College of Pharmaceutical Science, Zhejiang University, Hangzhou, People's Republic of China.
- 3State Key Laboratory of Natural Medicines, Key Laboratory of Drug Metabolism, China Pharmaceutical University, Nanjing, People's Republic of China.
- 4Department of Pharmacy, Zhejiang Academy of Traditional Chinese Medicine, Tongde Hospital of Zhejiang Province, Hangzhou, People's Republic of China.
- 5Department of pharmacy, Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, People's Republic of China.
- 6Department of Pharmacy, Affiliated Xiaoshan Hospital, Hangzhou Normal University, Hangzhou, People's Republic of China.
Abstract
Background
The high variability in the pharmacokinetics of tigecycline may impact therapeutic efficacy, particularly in critically ill patients. This study introduces machine learning (ML) methods, aiming to clarify their effectiveness in individualized drug exposure prediction and determine the optimal modeling strategy.Methods
The least absolute shrinkage and selection operator (LASSO) method was used to screen the covariates. Four ML models, random forest, extreme gradient boosting, multilayer perceptron, and long short-term memory network (LSTM), were trained and tested using 751 concentrations from 98 patients in the pharmacokinetic dataset to predict the area under the steady-state plasma concentration-time curve within 24 hours (AUC0-24h) of tigecycline. Furthermore, on the internal test set, we evaluated the accuracy of the published tigecycline Population Pharmacokinetics (PopPK) models.Results
Seven covariates were screened out by LASSO method, including total bilirubin, creatinine clearance, Sequential Organ Failure Assessment (SOFA) score, body mass index, serum creatinine, age, and prothrombin time. Internal validation showed that, among the four ML models, the LSTM model performed the best, with the combination of covariates and the plasma concentration feature "C₀ + C0.5 + C₆" exhibited the optimal predictive performance (R²: 0.887, RMSE: 2.496 mg·h/L, F₂₀: 75%). In comparison, the model combining covariates with C₀ was the most concise and better suited to practical computational conditions; its LSTM model was comparable to some PopPK models in terms of overall trend fitting and systematic bias control (R²: 0.818, RMSE: 3.164 mg·h/L, F₂₀ = 60%).Conclusion
The LSTM model established in this study enables an accurate estimate of tigecycline AUC0-24h. Three-point sampling provided the lowest prediction error, whereas C₀ alone reduced the sampling burden. These findings provide an effective modeling reference for tigecycline therapeutic drug monitoring.
Related
- High-Dose Tigecycline for Multidrug-Resistant Serratia Infections in Critically Ill Adults: A Systematic Review of Direct and Indirect Clinical Evidence.
- Development and validation of machine learning models for predicting pancreatic injury observed after tigecycline initiation: a retrospective real-world study
- Application of machine learning to identify risk factors for bloodstream infections caused by different antimicrobial resistance phenotypes in chemotherapy patients.
- Physiology-Guided Management of Acute Right Ventricular Failure and Pulmonary Hypertension in Critically Ill Patients: Hemodynamic Phenotypes, Ventilator Interactions, and Rescue Strategies
- Autonomic nervous system dysfunction in critically ill cancer patients: Hemodynamic consequences and multimodal assessment strategies☆
BibTeX
@article{dai2026application,
title = {{Application of Machine Learning Models to Predict Tigecycline Exposure in Critically Ill Patients Using Different Limited Sampling Strategies.}},
author = {Anran Dai and Hao Liu and Bo Yang and Feiyue Zheng and Yan Guan and Jieqiong Liu and Jianping Zhu and Gang Liang and Yi Yang and Lingyan Yu and Yuhua Zhao and Zhenwei Yu},
journal = {Drug design, development and therapy},
year = {2026},
month = oct,
doi = {10.2147/dddt.s630414},
url = {https://europepmc.org/article/MED/42894723},
note = {CC BY-NC},
}
RIS
TY - JOUR
TI - Application of Machine Learning Models to Predict Tigecycline Exposure in Critically Ill Patients Using Different Limited Sampling Strategies.
AU - Anran Dai
AU - Hao Liu
AU - Bo Yang
AU - Feiyue Zheng
AU - Yan Guan
AU - Jieqiong Liu
AU - Jianping Zhu
AU - Gang Liang
AU - Yi Yang
AU - Lingyan Yu
AU - Yuhua Zhao
AU - Zhenwei Yu
PY - 2026
DA - 2026/10/14
JO - Drug design, development and therapy
DO - 10.2147/dddt.s630414
UR - https://europepmc.org/article/MED/42894723
AB - Background The high variability in the pharmacokinetics of tigecycline may impact therapeutic efficacy, particularly in critically ill patients. This study introduces machine learning (ML) methods, aiming to clarify their effectiveness in individualized drug exposure prediction and determine the optimal modeling strategy.Methods The least absolute shrinkage and selection operator (LASSO) method was used to screen the covariates. Four ML models, random forest, extreme gradient boosting, multilayer perceptron, and long short-term memory network (LSTM), were trained and tested using 751 concentrations from 98 patients in the pharmacokinetic dataset to predict the area under the steady-state plasma concentration-time curve within 24 hours (AUC0-24h) of tigecycline. Furthermore, on the internal test set, we evaluated the accuracy of the published tigecycline Population Pharmacokinetics (PopPK) models.Results Seven covariates were screened out by LASSO method, including total bilirubin, creatinine clearance, Sequential Organ Failure Assessment (SOFA) score, body mass index, serum creatinine, age, and prothrombin time. Internal validation showed that, among the four ML models, the LSTM model performed the best, with the combination of covariates and the plasma concentration feature "C₀ + C0.5 + C₆" exhibited the optimal predictive performance (R²: 0.887, RMSE: 2.496 mg·h/L, F₂₀: 75%). In comparison, the model combining covariates with C₀ was the most concise and better suited to practical computational conditions; its LSTM model was comparable to some PopPK models in terms of overall trend fitting and systematic bias control (R²: 0.818, RMSE: 3.164 mg·h/L, F₂₀ = 60%).Conclusion The LSTM model established in this study enables an accurate estimate of tigecycline AUC0-24h. Three-point sampling provided the lowest prediction error, whereas C₀ alone reduced the sampling burden. These findings provide an effective modeling reference for tigecycline therapeutic drug monitoring.
SN - 1177-8881
ER -