JMIR Publications

ABSTRACT

Background:

Tacrolimus is the backbone of immunosuppression in solid organ transplantation, requiring precise dosing due to its narrow therapeutic range. Maintaining therapeutic tacrolimus levels post-operatively is challenging due to diverse patient characteristics, donor organ factors, drug interactions, and evolving perioperative physiology.

Objective:

The goal of this project is to design a machine learning model to predict the next-day tacrolimus trough concentration (C0) and guide dosing to prevent persistent under- or over-dosing.

Methods:

Retrospective data from 1,597 adult kidney and liver transplant recipients at UC San Diego Health were used to develop a Long Short-Term Memory (LSTM) model to predict next-day tacrolimus C0 in an inpatient setting. Predictors included patient demographics, comorbidities, vital signs, laboratory parameters, ordered diet, and medications. Permutation feature importance was evaluated for the model. We further implemented a classification task to evaluate the model’s ability to identify underdosing, therapeutic dosing, and overdosing. Finally, we generated next day dose recommendations that would achieve tacrolimus C0 within the target ranges.

Results:

The LSTM model provided a mean absolute error (MAE) of 1.880 ng/mL when predicting next-day tacrolimus C0. Top predictive features included the recent tacrolimus concentrations, tacrolimus doses, transplant organ type, diet and interactive drugs. When predicting underdosing, therapeutic, overdosing using a 3-class classification task, the model achieved a micro F1 score of 0.653. For dose recommendation, the best clinical outcomes were achieved when the actual total daily dose closely aligned with the model's recommended dose (within 3mg).

Conclusions:

As one of the largest studies applying artificial intelligence to tacrolimus dosing, our LSTM model effectively predicts tacrolimus C0 and could potentially guide accurate dose recommendations. Further prospective studies are needed to evaluate the model's performance in real-world dose adjustments.