Mitigating propensity score model misspecification with multiply robust weights when leveraging external data.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Statistical Inference

Background:

• External data augmentation of randomized controlled trials (RCTs) requires effective covariate adjustment.
• Propensity score methods are common but vulnerable to model misspecification due to unknown treatment selection.
• Model misspecification can lead to biased estimates in Bayesian dynamic borrowing methods.

Purpose of the Study:

• To develop a robust Bayesian inference procedure for integrating external data into RCTs.
• To improve robustness against propensity score model misspecification.
• To incorporate multiply robust weights into informative power priors for enhanced covariate adjustment.

Main Methods:

• Proposed a Bayesian inference procedure integrating multiply robust weights into power priors.
• Specified a set of candidate propensity score models to derive multiply robust weights.
• Extended the approach to accommodate multiple external datasets.

Main Results:

• Simulation studies demonstrated desirable operating characteristics when a correct model was included.
• Achieved low bias and root mean squared error (RMSE).
• Maintained controlled type I error rates and high statistical power.

Conclusions:

• The proposed method offers a robust strategy for covariate adjustment using external data, especially when selecting a single propensity score model is challenging.
• This approach enhances the reliability of estimates from augmented RCTs.
• Facilitates more effective use of external data in clinical research.