Development of a predictive model for nephrotoxicity during tacrolimus treatment using machine learning methods

Area of Science:

• Pharmacology
• Nephrology
• Data Science

Background:

• Tacrolimus (TAC) is crucial for preventing organ transplant rejection and treating autoimmune diseases.
• Nephrotoxicity is a significant adverse event associated with TAC, often linked to supra-therapeutic drug concentrations.
• Current monitoring may not prevent nephrotoxicity in all patients, necessitating individualized therapeutic strategies.

Purpose of the Study:

• To develop and validate a predictive model for individualized nephrotoxicity risk in patients receiving tacrolimus.
• To improve patient outcomes by identifying individuals likely to experience adverse renal effects.

Main Methods:

• Retrospective analysis of patient data, including demographics, concomitant medications, and tacrolimus whole-blood concentrations.
• Definition of nephrotoxicity as a serum creatinine increase within 60 days of TAC initiation.
• Development and comparison of 13 machine learning models (e.g., SVM, Random Forest) against a conventional model based solely on TAC concentration.

Main Results:

• A Support Vector Machine (SVM) model demonstrated superior performance in predicting nephrotoxicity.
• The best-performing SVM model achieved a high F2 score of 0.750, significantly outperforming the conventional model (0.500).
• The model was constructed using data from 163 patients and validated on 41 patients, predominantly with inflammatory or autoimmune conditions.

Conclusions:

• A novel machine learning model effectively predicts tacrolimus-induced nephrotoxicity using comprehensive patient data.
• This predictive tool can aid clinicians in identifying high-risk patients before initiating tacrolimus treatment.
• Enabling individualized therapeutic drug monitoring strategies can help prevent nephrotoxicity and optimize patient care.