Purpose: Tacrolimus, a widely prescribed immunosuppressive drug, requires therapeutic drug monitoring due to its high interindividual variability in pharmacokinetics and narrow therapeutic index. Since tacrolimus plasma levels obtained during therapeutic drug monitoring are automatically deposited to electronic medical records (EHRs) system, the key data to build a population pharmacokinetic (PK) model are available in the EHR system. However, there are several challenges that need to be addressed in order to use this data for population PK modeling. The goal of this study is to address some of these challenges and develop a population PK model for tacrolimus in kidney transplant recipients using data extracted from EHRs and a DNA biobank without collecting additional data.
Methods: We extracted tacrolimus plasma levels collected from therapeutic drug monitoring as well as patients' clinical factors from the EHRs. Tacrolimus dosing information was also extracted using natural language processing. Genotype data including CYP3A5 rs776746 were obtained from Vanderbilt’s DNA biobank, BioVU, which contains linked deidentified EHR data. We developed an algorithm to preprocess the extracted data and construct the PK data in the required format for population PK modeling. With the constructed PK data, we performed population PK modeling for tacrolimus in kidney transplant recipients using nonlinear mixed-effects modeling approach.
Results: We successfully preprocessed the extracted data and constructed the PK data using an R algorithm, and a population PK model for tacrolimus was developed with this data. A one-compartment model with first-order absorption was selected as the base model that provided reasonable estimates of tacrolimus apparent clearance (CL/F) and apparent volume of distribution (V/F). The CYP3A5*3 loss-of-function SNP, rs776746 most significantly influenced CL/F and albumin was significantly associated with V/F.
Conclusion: Our study demonstrated that the data required for population PK modeling can be extracted from EHR-linked biobanks, and the extracted data can be used to develop a population PK model in the context of routine clinical practice. The PK analysis confirmed previous studies that have established CYP3A5 rs776746 is associated with tacrolimus clearance. The developed population PK model would be useful for tacrolimus dose optimization for kidney transplant recipients. The developed tools that can address general challenges in using EHR data would be useful to develop population PK models for other medications with various data resources.