Informative study designs to identify true parameter-covariate relationships.

Area of Science:

• Pharmacometrics
• Statistical Modeling
• Biostatistics

Background:

• Accurate covariate identification is crucial for robust pharmacokinetic/pharmacodynamic (PK/PD) models.
• Study design can significantly impact the ability to discern true covariate relationships from spurious ones.

Purpose of the Study:

• To investigate how different study designs influence the selection of the correct covariate model.
• To compare the probability of selecting a 'true' model (lean body weight on clearance) versus a 'false' model (total body weight (WT) on clearance).

Main Methods:

• Comparison of two covariate models: lean body weight vs. total body weight (WT) on clearance.
• Evaluation of study designs with WT as lognormally distributed (non-stratified) versus stratified into 3 equal strata.
• Analysis of the probability of selecting the 'True Model' across varying WT inclusion criteria.

Main Results:

• The probability of selecting the 'True Model' increased with a wider WT inclusion criterion.
• The stratified design consistently yielded a higher probability of selecting the 'True Model' compared to the non-stratified design.
• Stratification of WT improved the identification of the true parameter-covariate relationship.

Conclusions:

• Stratified study designs, coupled with a broad covariate range, enhance the identification of true parameter-covariate relationships.
• This approach is vital for developing models that can be reliably extrapolated to diverse populations, such as for dosing adjustments in obesity.
• Optimizing study design is key to building more accurate and generalizable PK/PD models.