Crossover designs with correlated observations

Area of Science:

• Biostatistics
• Clinical Trial Design
• Experimental Design

Background:

• Crossover designs are frequently employed in medical research to compare multiple treatments.
• Observations within subjects in crossover trials are often correlated, complicating analysis.
• Existing methods may not fully account for complex correlation structures.

Purpose of the Study:

• To develop methods for constructing optimal crossover designs when observations within subjects are correlated.
• To investigate the impact of correlation structure parameters on design optimality.
• To determine optimal group sizes for subjects receiving specific treatment sequences.

Main Methods:

• Construction of crossover designs considering intra-subject correlation.
• Derivation of locally optimal designs when the correlation structure is known.
• Development of optimum Bayesian crossover designs using known parameter distributions.
• Optimization of group sizes for balanced treatment sequence allocation.

Main Results:

• The study provides a framework for building robust crossover designs adaptable to known correlation structures.
• Locally optimal designs are derived for specific correlation models.
• Bayesian optimal designs are constructed for scenarios with known parameter distributions.
• Methods for determining optimal subject group sizes are presented.

Conclusions:

• The proposed methods enhance the efficiency and accuracy of crossover trials with correlated data.
• Accounting for correlation structure is crucial for optimal design construction.
• The findings offer practical guidance for designing comparative medical studies.
• Optimal group size determination improves resource allocation in clinical trials.