Infrastructure for quality control of handling of reference data in PBPK projects using PK-Sim

Abstract: PBPK modelling is used in drug development to predict how the drug isaffected by the body. These models are built on physiology and describeprocesses like absorption, distribution, metabolism and excretion.Simulations of these processes can take place in the software PK-Sim forinstance. In PK-Sim, datasets of clinical observations can be importedfor model validation. When building the models, correct data and correcttype of model has to used to prevent incorrect predictions. These errorscan be discovered by performing quality control, preferably beforeinitiating model development in order to save time. An automatic processof quality control is also a way to save time, as well as to setstandards and provide transparency. At Pharmetheus, where PBPK modellingin PK-Sim is performed, quality control is executed manually. In thisproject, an infrastructure for quality control of data in PBPK projectsin PK-Sim was developed in R. The intention was to use thisinfrastructure before initiating model development and as a complementto manual quality control. The quality control was focused on dataintegrity and data concordance. Controls that are performed includedcorrectness of values and units for the observed data as well asensuring that each data had been assigned to the corresponding model.The developed model achieved to capture all defined errors to a highdegree. However, in many scenarios rounding of values occurred which wasnot always handled as intended. More evaluation also has to be performedbefore this infrastructure can be used in production.