TDMxR: an open-source package for model-based therapeutic drug monitoring in R
Sebastian G. Wicha
Dept. of Clinical Pharmacy, Institute of Pharmacy, University of Hamburg, Hamburg, Germany
Objectives: Pharmacometric models have gained popularity to support therapeutic drug monitoring (TDM) by model-based techniques such as ‘Probabilisting Dosing’, i.e. a priori prediction of a likely effective dose considering the covariates of the patient, and ‘Bayesian forecasting’, i.e. a posteriori prediction of an individual dose using patient covariates and previous TDM measurements [1]. We aimed to develop an R package to facilitate model-based TDM using state-of-the-art pharmacometric techniques.
Methods: TDMxR was developed under R version 3.5.2. The R package ‘mrgsolve’ [2] was utilized to provide an efficient simulation framework for pharmacometric models. The R package ‘data.table’ [3] provided an efficient framework to handle large-scale datasets originating from the simulation output. ‘doParallel’ [4] was utilised for optional parallelization.
Results: TDMxR provides an efficient framework for typical tasks in model-based TDM developed to maximize computing performance. The following options are available:
- Flexible handling of complex dosing regimens
- Performance of stochastic simulations
- Bayesian estimation of individual pharmacokinetic parameters
- Handling of inter-occasion variability incl. weighting functions.
- Simulation from the posterior distribution to visualize uncertainty of the estimated PK or PD profile.
- Probability of target attainment calculation from a priori and a posteriori simulations for user-defined endpoints for PK or PD.
- Deterministic and probabilistic dose optimisation based on user-defined PK or PD targets
The results were successfully cross-validated against NONMEM 7.4.1 and individual parameters estimated by TDMxR were unbiased (absolute mean bias <0.1%) and varied numerically by
Conclusions: TDMxR provides an efficient framework for model-based TDM providing state-of-the-art functionality at optimised computational cost. The R-Shiny-based TDMx software [1] is currently updated to be operated by TDMxR. Further development will include interfaces to other simulation packages and inclusion of an optimal sampling module. A release of the TDMxR package to CRAN is planned.
References:
[1] Wicha SG, Kees MG, Solms A, Minichmayr IK, Kratzer A, Kloft C. TDMx: A novel web-based open-access support tool for optimising antimicrobial dosing regimens in clinical routine. Int J Antimicrob Agents 2015;45. doi:10.1016/j.ijantimicag.2014.12.010.
[2] https://CRAN.R-project.org/package=mrgsolve
[3] https://CRAN.R-project.org/package=data.table
[4] https://CRAN.R-project.org/package=doParallel