Model qualification is more popularly known as model "validation." The word "validation" implies a procedure of utmost robustness and may not be applicable to the usual PK/PD models that are found in the literature. Further, the fact that the true model and its parameters are not known makes the choice of the word "validation" even poorer. A contrasting example would be the validation of an analytical method, where "true" concentrations of the chemical entity are known for making a calibration standard. For wider acceptance, all models are required to be qualified and credible. Clear specification of the purpose for which the model is being developed is a prerequisite for any model building exercise.
Qualified Model/parameters. A model and its set of parameters are deemed "qualified" to perform particular task(s) if they satisfy prespecified criteria. Example: Application of posterior predictive check to a model and its parameters for use in Monte Carlo simulations [4, 5].
Credible Model/parameters. A model and its set of parameters are deemed "credible"  to perform particular task(s) if the conceptual foundation on which the model was proposed is satisfactory to a group of experts (subject matter-experts). Although there is no formal record of the existence of such models, to the best of our knowledge, we speculate that (at least the structural) models for warfarin  and reverse transcriptase/ aspartyl protease inhibitors  would be deemed as "credible."
Monte Carlo simulations can be used to qualify a given model and its parameters. Based on the objective, qualification methods can test either the descriptive capacity or the extrapolation capacity of a given model. Adequate description of the data will ensure that the proposed model and its parameters are qualified to make inferences reliably within the range of the data studied. Such a qualification will be assessed using the routine diagnostic tests such as plots of the independent variable vs. observed and (individual/population) predicted, summary statistics and determining the precision of the parameter estimates. For example, developing an acceptable descriptive model is critical for making labeling recommendations. Product labels, usually, do not extrapolate results beyond the data range observed. A model is qualified to predict beyond the range of the data used for building the model if the descriptive capacity of the model is acceptable and the model (and parameters, if applicable) is credible. It is important to note that there is no means of assessing whether a model can be used for extrapolation. Hence the credibility of the model i.e., whether the model was derived from sound physiological principles and whether the submodel and its parameters appear reasonable to a panel of experts, is important.
The guidance for industry on population pharmacokinetics presents a variety of simulation methods that can be used to "qualify" models/ parameters . Although a variety of methods for model qualification are known, no thorough evaluation of their advantages and disadvantages is available.
Was this article helpful?