Identifiability: Its Role in Design of Pharmacokinetic Experi ments

Identifiability analysis is a vital first step in the design of experiments to estimate phannacokinetic parameters; it can be of help in choosing the model structure, the numbers and locations of inputs and outputs, and the input-signal waveforms; it can also be of help in deciding whether certain physical parameters need to be measured directly, and whether the gains of certain measurement transducers need to be calibrated. The effect of such choices and decisions on the identifiability of the model can be tested by standard algebraic procedures. These provide answers to questions of nonexistence of a solution (underidentification), multiple solutions (local identification), and redundant equations (overidentification). The practical implications of identiflability are demonstrated through pharmacokinetic examples, but the results are equally applicable in other fields.