Blinded sample size re-estimation for recurrent event data with time trends

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Traditional fixed sample size designs have limitations in managing uncertainty.
• Existing blinded sample size re-estimation (BSSR) methods for recurrent events assume constant event rates, which is often unrealistic.
• Relapsing multiple sclerosis studies often exhibit time trends in event rates.

Purpose of the Study:

• To propose and evaluate methods for BSSR in recurrent event data when event rates have a time trend.
• To address limitations of current BSSR approaches in complex clinical scenarios.
• To ensure robust sample size planning and accurate final analysis in clinical trials.

Main Methods:

• Development of BSSR methods based on a proportional intensity frailty model to account for time trends.
• Utilizing standard negative binomial methods for initial sample size planning and final analysis.
• Employing a full likelihood analysis for interim sample size re-estimation.
• Conducting a simulation study motivated by relapsing multiple sclerosis data.

Main Results:

• The proposed BSSR procedure effectively controls the type I error rate.
• The method maintains desired statistical power even with misspecified nuisance parameters.
• Study duration and recruitment period length significantly impact operating characteristics.
• The approach is suitable when patient follow-up time is balanced across treatment groups.

Conclusions:

• The proposed BSSR methods offer a more accurate approach for sample size determination in recurrent event studies with time trends.
• This methodology enhances the reliability of clinical trial results, particularly in diseases like multiple sclerosis.
• The findings provide valuable guidance for optimizing clinical trial design and statistical analysis.