Tacrolimus Exposure Prediction Using Machine Learning.

Area of Science:

• Pharmacokinetics and Pharmacodynamics
• Machine Learning in Medicine
• Transplant Medicine

Background:

• Tacrolimus (TAC) is a vital immunosuppressant post-transplantation.
• Accurate estimation of TAC's area under the curve (AUC) is crucial for therapeutic drug monitoring and dose adjustment.
• Current methods may be complex or time-consuming for routine clinical use.

Purpose of the Study:

• To develop and validate machine learning (ML) models for estimating tacrolimus (TAC) interdose area under the curve (AUC).
• To compare the performance of ML models against traditional Bayesian estimation methods.
• To assess the utility of ML models for routine TAC exposure estimation and dose adjustment in organ transplant recipients.

Main Methods:

• Utilized 4,997 (b.i.d.) and 1,452 (q.d.) TAC AUC data points from an expert system.
• Developed four Xgboost ML models using 2 or 3 TAC concentrations and covariates (dose, transplant type, age, time since transplant).
• Validated models using data splitting (75% train, 25% test), 10-fold cross-validation, and six independent full-PK datasets.

Main Results:

• Xgboost ML models achieved excellent AUC estimation performance in test datasets (relative bias <5%, relative RMSE <10%).
• Models demonstrated superior performance compared to maximum a posteriori Bayesian estimation in independent full-PK datasets.
• The developed models accurately estimate TAC interdose AUC using limited concentration data.

Conclusions:

• Xgboost ML models provide accurate and efficient estimation of TAC interdose AUC.
• These models can be implemented for routine clinical monitoring and dose adjustment of tacrolimus.
• The findings support the integration of ML into clinical practice for optimizing immunosuppressive therapy.