PD and Reaction Network Modeling

Pharmacodynamics (PD) describes the interactions of drugs with the organisms.

This includes binding of drugs to their targets (which may also be relevant for understanding the PK) and resulting direct or indirect effects. Pharmacodynamics can also relate drug concentration profiles (PK) to clinical endpoints, which usually requires consideration of disease progression. Various modeling approaches are used to analyze PD or disease progression, either alone or in combination with PK. Simple models use a sigmoidal function to relate a concentration to its effects (e.g., Hill, IC50, Emax-shape). Here, the maximum effect and the concentration that corresponds to half the maximum effect are typical curve characteristics. Such approaches can be combined with advanced statistical methods, as the effect of a drug is rarely fully deterministic. Inter-individual or inter-occasion differences can be investigated and quantified in such a way, see Modeling Concepts - PK and PD Modeling.

Transferring or translating such models to new application scenarios is often not straightforward. A typical question to answer would be how the curve characteristics are expected to change in the new situation. However, if the model explicitly considers the crucial physiologic and mechanistic aspects, and it is known from independent experiments how these change in certain scenarios, a translation and prediction can become feasible. The level of mechanistic detail needed depends on the particular problem. In certain cases, the desired detail might include sophisticated reaction networks. Sometimes, these are sufficient on their own to understand relevant PD behavior. Detailed pathway modeling is also a major activity in systems biology in academia. Many excellent reviews are devoted to this topic, which will not be further detailed here.

In other cases or for other questions, the PK/PD interaction is important to consider. Here, PBPK models offer an intuitive framework for coupling PK with simple or mechanistic PD models. See [18] for a recent example.