Design and analysis of individually randomized multiple baseline factorial trials

Area of Science:

• Behavioral Science
• Rare Disease Research
• Clinical Trial Design

Background:

• Assessing behavioral interventions in rare diseases is difficult due to small sample sizes and ethical constraints.
• The standard multiple baseline design (MBD) is limited to evaluating single interventions.
• Factorial designs are often infeasible in rare diseases due to participant limitations.

Purpose of the Study:

• To propose and describe the individually randomized multiple baseline factorial design (MBFD) for evaluating multiple interventions in rare diseases.
• To clarify estimands and introduce statistical models for analyzing MBFD data.
• To assess the statistical performance of MBFD through simulations.

Main Methods:

• Description of the standard MBFD characteristics.
• Introduction of three statistical models (LMM, GEE) for data analysis.
• Simulation studies to compare model biases, sizes, and power.

Main Results:

• The MBFD requires fewer participants than standard factorial designs while maintaining statistical power.
• Generalized estimating equations (GEE) are recommended over linear mixed effect models (LMM) to address potential random effect misspecifications.
• Small sample corrections (e.g., Mancl and DeRouen variance estimator) are suggested for sample sizes under 120.

Conclusions:

• The MBFD is a viable and statistically robust design for rare disease intervention research.
• GEE with small sample corrections offers reliable analysis for MBFD data.
• This design enhances the ability to evaluate multiple interventions and their combinations in rare disease populations.