Dynamic Borrowing With a Bias-Tolerance Cap in Augmented Randomized Controlled Trials.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Epidemiology

Background:

• Randomized controlled trials (RCTs) are the gold standard but face challenges like cost and duration.
• External control data offers a solution but is susceptible to confounding bias from unmeasured covariates.
• Existing methods often fail to adequately address or quantify this confounding bias.

Purpose of the Study:

• To propose a novel statistical method for estimating confounding effects of unmeasured covariates when using external control data.
• To develop an estimator that dynamically incorporates external data while adjusting for confounding.
• To control the expected bias within a prespecified tolerance cap for better stakeholder discussions.

Main Methods:

• Utilized model-based regression standardization, inverse probability weighting, and augmented inverse probability weighting for continuous/binary outcomes.
• Developed a dynamic borrowing estimator using a weighted mean, adjusting weights based on estimated confounding.
• Incorporated a 'bias-tolerance cap' to manage the impact of unmeasured confounding.

Main Results:

• The proposed method effectively regulated bias within the specified tolerance cap, irrespective of confounding magnitude.
• Demonstrated significant improvements in statistical power and efficiency when confounding was absent.
• Successfully applied the method to a real-world case involving advanced pancreatic cancer.

Conclusions:

• The novel method provides a robust approach to mitigate confounding bias when integrating external control data into RCTs.
• The bias-tolerance cap facilitates informed decision-making regarding the acceptability of bias in effect estimates.
• This approach enhances the utility of external data, potentially reducing trial costs and duration while maintaining statistical rigor.