The B-value: a tool for monitoring data

Area of Science:

• Biostatistics
• Clinical Trial Monitoring
• Statistical Methods

Background:

• Monitoring clinical trial data, especially with non-sequential designs, presents challenges for accurate interim analysis.
• Traditional Z-values are often misinterpreted and do not allow for simple projections of study outcomes.
• Slowly accruing data complicates the assessment of conditional power and future study trends.

Purpose of the Study:

• To introduce and describe the utility of the B-value for monitoring non-sequential experiments.
• To provide a method for calculating conditional power using transformed Z-values.
• To enable reliable extrapolations of study trends under various hypotheses.

Main Methods:

• The paper proposes the use of a B-value, a transformation of the Z-value.
• Conditional power is calculated using this B-value.
• Observed data trends are modeled as a Brownian motion process for extrapolation.

Main Results:

• The B-value and its decomposition facilitate simple extrapolations to the study's end.
• This method addresses the limitations of interim Z-values in non-sequential designs.
• Demonstrated applications across various statistical tests, including Z-tests and the log-rank test.

Conclusions:

• The B-value offers a robust approach for interim data monitoring in clinical trials.
• It improves the interpretability and predictive accuracy of study results.
• The method is applicable to a range of common statistical tests used in research.