A pharmacokinetic model can reproduce observed concentrations while leaving its performance in new circumstances open to testing. A 2018 study developed an agent-based model of human oral absorption, using midazolam and clonazepam as examples.
The in silico drug absorption tract, or ISDAT, represented physiological spaces, membranes, and metabolic processes. Simulations were compared with reference clinical concentration–time profiles using visual agreement, weighted error measures, and differences in key pharmacokinetic parameters. Model structure and parameters were iteratively refined until the specified criteria were met.
Both drug simulations met the study’s similarity criteria. The model also predicted absolute oral bioavailability of 44% for midazolam and 93% for clonazepam, values the authors compared with published ranges. Those bioavailability estimates were simulation outputs. They were distinct from the clinical measurements used as reference data.
Agreement after iterative refinement demonstrates successful reproduction of the selected reference behavior. Confidence in a new application also depends on the population, formulation, coexposures, and conditions being simulated. A reviewer can distinguish the data used to refine a model from data used to challenge a subsequent prediction. The chosen similarity thresholds define what counted as success in this paper. For an individual exposure reconstruction, the model’s assumptions and uncertainty need to accompany any predicted concentration or absorbed fraction. The study supplies a modeling approach and documented fit, with wider applicability requiring evidence suited to the proposed use.
Read the full scientific paper. DOI 10.1371/journal.pone.0203361.