Saddlepoint approximation for weighted log-rank tests based on block truncated binomial design

Area of Science:

• Biostatistics
• Clinical Trials
• Survival Analysis

Background:

• Clustered data are common in biomedical research and clinical trials.
• Log-rank tests are standard for comparing two independent samples of clustered data.
• Randomized block and truncated binomial designs are used to balance and reduce bias.

Purpose of the Study:

• To approximate p-values for log-rank tests on clustered survival data.
• To evaluate the accuracy of the double saddlepoint approximation method.

Main Methods:

• Survival data were randomized using a generalized randomized block design.
• Clustered data within blocks were randomized by a truncated binomial design.
• The double saddlepoint approximation method was employed for p-value approximation.

Main Results:

• The double saddlepoint approximation accurately approximates p-values for the null permutation distribution of log-rank tests.
• Numerical studies confirm the high accuracy of this approximation.
• The proposed method demonstrates superior accuracy compared to asymptotic normal approximation.

Conclusions:

• The double saddlepoint approximation is a highly accurate method for analyzing clustered survival data.
• This technique enhances the analysis of log-rank tests in clinical trials with clustered data.
• The findings suggest a more reliable approach for statistical inference in such settings.