Case weighted power priors for hybrid control analyses with time-to-event data

Area of Science:

• Biostatistics
• Clinical Trial Design
• Health Economics

Background:

• Randomized controlled trials (RCTs) are the gold standard for evaluating treatment efficacy.
• Internal control arms in RCTs can be limited by sample size and patient characteristics.
• External control data offers potential to augment RCTs but requires careful handling of systematic differences.

Purpose of the Study:

• To develop a novel method for hybrid analyses combining internal and external control data in RCTs.
• To enable flexible borrowing of information from external controls based on patient similarity.
• To provide robust statistical inference when external data is not perfectly compatible with RCT data.

Main Methods:

• A power prior method is extended to compute data-dependent discounting weights for each external control.
• Weights are determined by assessing compatibility between external control patients and RCT data using predictive distributions.
• A proportional hazards regression model with piecewise constant baseline hazard is employed for time-to-event analysis.

Main Results:

• The proposed case-weighted power prior method demonstrated robust inference in simulations and a real-world non-small cell lung cancer trial.
• The method effectively accounts for systematic differences, including unmeasured confounders, between RCT and external control populations.
• Adjustable borrowing strength ensures reliable augmentation of internal control arms.

Conclusions:

• The hybrid analysis method offers a statistically sound approach to leverage external control data in RCTs.
• This methodology enhances statistical power and efficiency in clinical trial evaluations.
• The approach is particularly valuable when internal control groups are small or when external data sources are available.