Simulation-based power calculations for planning a two-stage individual participant data meta-analysis.

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trials

Background:

• Individual Participant Data (IPD) meta-analyses are resource-intensive.
• Assessing statistical power is vital for efficient planning and funding of IPD projects.
• A simulation-based approach offers a robust method for power calculations.

Purpose of the Study:

• To propose and illustrate a simulation-based power calculation framework for planned IPD meta-analyses.
• To estimate the statistical power for detecting treatment-covariate interactions in continuous outcomes.
• To guide researchers and funders in optimizing IPD meta-analysis projects.

Main Methods:

• A four-step simulation process: specifying a data-generating model, defining parameters from existing data, simulating IPD meta-analysis datasets, and repeating simulations to estimate power.
• Utilizing a two-stage IPD meta-analysis framework to obtain summary estimates and p-values.
• Applying the method to a planned IPD meta-analysis of lifestyle interventions for weight gain in pregnancy, examining treatment-BMI interactions.

Main Results:

• A planned IPD meta-analysis had <60% power to detect a 1kg weight gain reduction per 10-unit BMI increase.
• Including IPD from 10 additional trials could increase power to over 80% under fixed-effect assumptions.
• Incorrectly dichotomizing BMI or discarding data significantly reduced statistical power.

Conclusions:

• Simulation-based power calculations are essential for planning and funding IPD meta-analysis projects.
• Routine use of this methodology can enhance the efficiency and reliability of IPD research.
• Informing decisions on data acquisition and analytical strategies improves the likelihood of detecting meaningful treatment effects.