Blinded sample size re-estimation for comparing over-dispersed count data incorporating follow-up lengths

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Blinded sample size re-estimation (BSSR) is crucial for maintaining statistical power in clinical trials.
• Conventional BSSR methods may fail for overdispersed count data if the working variance function is misspecified.
• Misspecification of the variance function can lead to reduced power in comparative clinical trials.

Purpose of the Study:

• To propose a novel BSSR method robust to working variance function misspecification.
• To ensure reliable power recovery in clinical trials with overdispersed count data.
• To address limitations of existing BSSR methods in adaptive trial designs.

Main Methods:

• Developed a weighted estimator for the dispersion parameter in BSSR.
• Incorporated weights to account for differences in follow-up length distributions.
• Utilized simulation studies to evaluate method performance.

Main Results:

• The proposed BSSR method demonstrated stable statistical power under variance function misspecifications.
• The weighted estimator mitigated power loss associated with incorrect variance function assumptions.
• The method was applied to a hypothetical COPD exacerbation trial.

Conclusions:

• The proposed BSSR method offers enhanced robustness for clinical trial sample size re-estimation.
• This adaptive design improves reliability when analyzing overdispersed count data.
• The findings have implications for optimizing clinical trial efficiency and accuracy.