Design of Experiment (DOE) Methods Maximize Information from a Minimal Number of Animals in Special Cases of Preclinical Bioavailability Testing

The use of statistical design techniques in various scientific fields has increased in recent years fostered by the broad availability of sophisticated design of experiment (DOE) software. In the 1970s, Sucker and Leuenberger developed some pioneering pharmaceutical applications. More recently, DOE has been used for solid dosage forms, – 6] drug solutions, emulsions, liposomes, and liquid-filled capsules. Its application to the avoidance of process problems was recently reviewed. However, there are few, if any, instances in the literature in which DOE has been applied to an animal study comparing the pharmacokinetics of different drug formulations. Most industrial pharmacokinetic studies follow a conventional design, varying only one factor at a time (Figure 1). Thus factor A may represent different dose levels (coded 1 and +1) studied in two groups of four animals. If we then wish to investigate another factor, e.g., the effect of food, we have to decide which dose level to leave constant. But no matter what our choice, we cannot estimate the food effect at other dose levels from this type of study design. Another drawback is that the number of animals increases considerably with the number of factors to be studied, easily resulting in a multiplicity of animal trials split into several blocks. A better approach is to accept that we usually need to study a number of factors for any drug with problematic biopharmaceutics. Pharmacokinetic studies are thus best planned at the start of a project, using DOE techniques such as screening designs or factor influence studies, in which the different factors are simultaneously varied. Such experiments are randomized and conducted in as few blocks as possible. All are then used to evaluate an effect, giving optimal precision to the resulting estimate. The overall advantage of DOE is that it reduces the number of animal tests while increasing the quality of the information generated, by quantifying additional factors and interactions and— usually—increasing the statistical power. This enables the study to provide the required information at an acceptable level of confidence. There can also be crucial logistic advantages for the formulation scientist in a more blockwise conduct of pharmacokinetic studies: increasing the number of parallel animal tests, assuming it is feasible to do so, is usually faster than consecutive studies. A large well-planned block of animal tests is also more readily outsourced, which is