Simultaneous versus sequential optimal design for pharmacokinetic-pharmacodynamic models with FO and FOCE...

Area of Science:

• Pharmacometrics
• Nonlinear Mixed Effects Models
• Pharmacokinetic-Pharmacodynamic (PKPD) Modeling

Background:

• Nested multiple response models are crucial in pharmacometrics but sensitive to methodological assumptions.
• Sequential and First Order (FO) estimation techniques can introduce bias in these models.
• Understanding design consequences of different estimation methods (FO vs. First Order Conditional Estimation (FOCE)) and design strategies (sequential vs. simultaneous) is critical.

Purpose of the Study:

• To investigate the impact of different design strategies and estimation methods on nested multiple response models.
• To evaluate the influence of nonlinearity on optimal design choices.
• To compare predicted standard errors with empirical estimates from simulation studies.

Main Methods:

• Developed design theory for nested multiple response models.
• Compared sequential versus simultaneous design strategies with FO and FOCE.
• Incorporated parameter-effects curvature (nonlinearity) into the design optimization.
• Utilized a pharmacokinetic-pharmacodynamic model for investigation.
• Extended methodology to include discrete (binary) response variables.
• Validated findings using simulation/estimation studies in NONMEM.

Main Results:

• Sequential and FO approaches demonstrated potential for biased results, particularly under higher nonlinearity.
• Simultaneous design with FOCE generally provided more robust and accurate parameter estimation.
• The study explored designs for continuous and discrete response variables.
• For binary response models, sequential optimization (product design) may offer near-optimal results due to limited information content.

Conclusions:

• Simultaneous design and FOCE are preferable for nested multiple response models to mitigate bias.
• Nonlinearity significantly influences optimal design choices, necessitating its inclusion in the optimization process.
• Discrete responses in PKPD models may require specific design considerations, potentially favoring sequential optimization strategies.