Some new confidence intervals for Kaplan-Meier based estimators from one and two sample survival data.

Area of Science:

• Biostatistics
• Clinical Trials
• Survival Analysis

Background:

• Restricted mean survival time (RMST) is crucial for evaluating treatment effects in survival studies.
• Traditional Greenwood's formula for RMST variance yields liberal confidence intervals (CIs) in small to moderate sample sizes.

Purpose of the Study:

• To develop novel, more accurate confidence intervals for RMST in single and two-sample survival analyses.
• To enhance interval estimation for comparing milestone survival probabilities.

Main Methods:

• Proposed empirical likelihood ratio, score-type, and log-log transformed CIs for single-sample RMST.
• Utilized variance estimates recovery technique for two-sample RMST difference and ratio CIs.
• Developed a new variance estimate for two-group comparisons, accommodating different censoring rates.

Main Results:

• The proposed CIs demonstrate improved performance for RMST estimation.
• The new variance estimate is effective for superiority and noninferiority testing, particularly when survival curves are similar.
• Methods for milestone survival probability comparisons also show good performance.

Conclusions:

• The novel confidence interval methods provide more reliable statistical inference for RMST and milestone probabilities.
• These advancements are particularly beneficial for clinical trials with limited sample sizes or closely related survival curves.