Analyzing survival curves at a fixed point in time.

Area of Science:

• Biostatistics
• Medical Statistics
• Survival Analysis

Background:

• Comparing survival curves is common in medical research.
• Focus is often on survival probability at a fixed time point.
• The naive test using survival estimates is frequently employed.

Purpose of the Study:

• To examine the performance of alternative statistical tests for comparing survival curves at a fixed time point.
• To compare type I errors and power of various tests via Monte Carlo simulation.
• To explore extensions for stratified data and regression analysis.

Main Methods:

• Evaluation of tests based on survival function transformations.
• Application of a generalized linear model for pseudo-observations.
• Monte Carlo simulation to assess type I errors and power across sample sizes.

Main Results:

• Performance of alternative tests was compared against the naive approach.
• The study assessed the statistical power and error rates of different methods.
• Pseudo-value approach demonstrated applicability in detailed regression analysis.

Conclusions:

• Alternative statistical tests offer improved performance for comparing survival curves at a fixed time point.
• The pseudo-value approach provides a flexible framework for survival probability analysis.
• Methods are applicable to stratified data and complex regression models, as shown in bone marrow transplant survival studies.