Two-sample survival tests based on control arm summary statistics

Area of Science:

• Biostatistics
• Clinical Trial Analysis
• Survival Analysis

Background:

• The one-sample log-rank test is standard for single-arm survival trials, comparing patient outcomes to a reference curve.
• Classical tests assume the reference curve is known, ignoring potential sampling errors from historical data.
• Ignoring reference curve variability can inflate the type I error rate in survival analyses.

Purpose of the Study:

• To develop a new survival test that accounts for the sampling error of estimated historical reference curves.
• To enable valid historical comparisons in single-arm trials when only a survival curve, not full historical data, is available.
• To provide a method applicable when the two-sample log-rank test is not feasible due to data limitations.

Main Methods:

• Proposed a novel survival test addressing the sampling error of reference survival curves.
• Developed sample size calculation formulas for the new test.
• Conducted a simulation study to evaluate the new test's performance.

Main Results:

• The new test effectively accounts for the sampling error of the reference curve.
• Demonstrated the test's validity even when only a historical survival curve is available.
• Simulation results showed the proposed method controls type I error rates.

Conclusions:

• The new survival test offers a valid approach for historical comparisons in single-arm trials.
• This method is crucial when individual historical patient data is unavailable.
• The developed formulas and validated test enhance the reliability of survival trial outcomes.