Development of in Vitro-in Vivo Correlations Using Various Artificial Neural Network Configurations

Abstract

It is desirable to have a predictive tool to determine the in vivo pharmacokinetics based on the in vitro dissolution and other important variables. We can see the in vitro — in vivo correlation (IVIVC) as an input-output relationship, and often are not interested in the internal structure of this model as long as we have a good, validated, predictive tool. This may be important, for example, in product development or to establish dissolution specifications. Many of the previous examples in this book use parametric models to define an IVIVC. For example, simple linear models are often used to relate a parameter or a time point descriptive of the dissolution to a parameter or a time point descriptive of the pharmacokinetic absorption^1–3. These models, however, can be unsuccessful in completely describing the IVIVC, and sometimes no correlation can be determined. The number of possible variables, the model unable to account for some physiological rate determining process, and the possible amount of variability intrinsic to the parameters of these modeled relationships are some examples of these difficulties ^4–6.