Conditional analysis of mixed Poisson processes with baseline counts: implications for trial design and analysis

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Modeling

Background:

• Clinical trial design often relies on marginal comparisons of treatment responses.
• Gaussian models have shown efficiency gains using baseline response data.
• Limited methods exist for count or point process data with baseline measurements.

Purpose of the Study:

• To present methods for designing and analyzing clinical trials with count or point process responses and baseline data.
• To evaluate the impact of selection criteria on power and sample size.
• To offer a more efficient alternative to existing designs.

Main Methods:

• Development of a conditional negative binomial model for response given baseline count.
• Application of the model to assess power and sample size requirements.
• Comparison of proposed methods with existing approaches.

Main Results:

• The conditional negative binomial model provides a framework for analyzing count/point process data with baseline.
• The proposed design methods demonstrate increased efficiency.
• Selection criteria can be effectively evaluated for their impact on trial parameters.

Conclusions:

• Conditional modeling with baseline counts offers superior efficiency in clinical trial design for specific data types.
• The presented methods are advantageous over prior designs for count and point process outcomes.
• This approach enhances statistical power and optimizes sample size determination.