How to choose a time zero for patients in external control arms

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Single-arm oncology trials often use external control arms from retrospective registries for efficacy assessment.
• Comparing novel compounds to physician's choice of chemotherapy presents challenges in selecting appropriate external control arm patients.
• Patients in retrospective registries may qualify for external control arms at multiple treatment lines.

Purpose of the Study:

• To investigate bias in external control arm construction for oncology trials.
• To evaluate methods for selecting patients for external control arms based on treatment lines.
• To propose an unbiased method for constructing external control arms using retrospective data.

Main Methods:

• Probabilistic arguments and simulation studies were employed.
• Real-world data was utilized for simulations.
• Comparison of overall survival between single-arm trial and various external control arm construction methods.

Main Results:

• Including external control arm patients from the last available line of therapy introduces bias.
• Randomly selecting a line of therapy for external control arm patients also results in bias.
• Bias favoring the single-arm trial was observed in both tested methods.

Conclusions:

• Current methods for constructing external control arms in oncology trials can lead to biased efficacy assessments.
• An unbiased approach involves including external control arm patients multiple times, corresponding to each qualifying treatment line.
• This multi-inclusion strategy enhances the reliability of overall survival comparisons in single-arm oncology studies.