Increasing the sample size when the unblinded interim result is promising.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Increasing sample size based on interim results can inflate Type I error rates.
• Existing methods include early stopping for futility, modifying test statistics, adjusting critical values, or enforcing sample size recalculation rules.

Purpose of the Study:

• To demonstrate that increasing sample size with promising unblinded interim results does not inflate Type I error.
• To establish criteria for a 'promising' interim result that obviates the need for statistical adjustments.

Main Methods:

• Review of existing statistical approaches for adaptive clinical trials.
• Development and application of a conditional power threshold (50%) to define a promising interim result.
• Simulation studies to evaluate Type I error control in group sequential trials with one or more interim analyses.

Main Results:

• Increasing sample size when conditional power exceeds 50% does not inflate Type I error.
• The 50% conditional power rule is effective for single interim analysis group sequential trials.
• Type I error is well-controlled in trials with multiple interim analyses under the proposed condition.

Conclusions:

• No statistical adjustments are necessary when increasing sample size based on promising unblinded interim results (conditional power > 50%).
• This finding simplifies adaptive clinical trial design, particularly for group sequential trials.
• The approach is robust for trials with multiple interim analyses.