Optimizing Tacrolimus Dosing During Hospitalization After Kidney Transplantation: A Comparative Model Analysis

Area of Science:

• Nephrology
• Pharmacology
• Biostatistics

Background:

• Optimizing tacrolimus dosing post-kidney transplant is crucial for preventing rejection and minimizing toxicity.
• Patient pharmacokinetic variability complicates early postoperative dose adjustments.
• Machine learning and statistical methods offer potential solutions for personalized dosing.

Purpose of the Study:

• To evaluate machine learning and statistical methods for optimizing tacrolimus dosing in early post-kidney transplant hospitalization.
• To compare the performance of Extreme Gradient Boosting (XGBoost), Elastic Net (EN), and Linear Regression (LR) models.

Main Methods:

• Retrospective analysis of 749 kidney transplant recipients' data (2015-2019).
• Collected tacrolimus doses, trough levels, and clinical variables during the first 12 postoperative days.
• Trained and validated models using 5-fold cross-validation, assessing performance with R² and RMSE.

Main Results:

• Elastic Net (EN) demonstrated superior performance with R² of 0.861±0.044 and RMSE of 0.930±0.220.
• Linear Regression (LR) and XGBoost provided clinically relevant predictions, though with slightly lower accuracy.
• External validation was not performed, indicating a need for further generalizability assessment.

Conclusions:

• Elastic Net regression is a practical and reliable model for predicting optimal tacrolimus dosage.
• Machine learning and statistical approaches are valuable tools for optimizing tacrolimus dosing in kidney transplant recipients.
• Multi-center validation is recommended for future studies to enhance clinical applicability.