Rapid sample size calculations for a defined likelihood ratio test-based power in mixed-effects models.

Area of Science:

• Pharmacometrics
• Statistical Modeling
• Computational Biology

Background:

• Traditional power calculations for Likelihood Ratio Test (LRT)-based hypothesis testing in mixed-effects models rely on computationally intensive simulations.
• Existing efficient methods are limited to Wald test-based inference.

Purpose of the Study:

• To introduce and validate the Monte Carlo Mapped Power (MCMP) method for efficient power and sample size calculations in mixed-effects models.
• To compare the MCMP method against traditional simulation-based approaches.

Main Methods:

• The MCMP method utilizes the difference in individual objective function values (ΔiOFV) from simulated datasets.
• Simulated data from a full model are re-estimated using both full and reduced models to derive ΔiOFVs.
• Power is determined by the percentage of summed ΔiOFVs (∑ΔiOFVs) exceeding a significance criterion across various sample sizes.

Main Results:

• The MCMP method demonstrated concordance with traditional power assessment methods.
• For achieving 90% power, the difference in required sample size was typically less than 10%.
• MCMP achieved relevant power information in less than 1% of the runtime of Simulation and Re-estimation (SSE) methods.

Conclusions:

• The MCMP method provides a fast and accurate approach for predicting power and sample size relationships.
• This method significantly reduces the computational burden associated with power calculations for LRT-based inference.
• MCMP is a valuable tool for optimizing study design in pharmacometric modeling.