Enhanced doubly robust estimation with concave link functions for estimands in clinical trials

Area of Science:

• Statistics
• Biostatistics
• Causal Inference

Background:

• Observational studies and non-randomized trials often exhibit baseline covariate imbalance between treatment and control groups.
• Traditional treatment effect estimates are biased under such imbalances, necessitating causal inference methods.
• Doubly robust estimators (DRE) offer a popular approach but rely on correct model specification, which is often challenging.

Purpose of the Study:

• To develop an enhanced doubly robust estimator (DRE) for unbiased treatment effect estimation.
• To address limitations of existing DREs by incorporating semiparametric models with concave link functions for propensity score and outcome models.
• To investigate the asymptotic properties and evaluate the performance of the proposed method.

Main Methods:

• Developed an enhanced doubly robust estimator utilizing semiparametric models.
• Employed concave link functions for both propensity score and outcome models.
• Studied the asymptotic properties of the proposed estimator and conducted simulation studies.

Main Results:

• The proposed enhanced doubly robust estimator demonstrates improved performance in simulation studies.
• The method provides a robust approach to estimating treatment effects even with potential model misspecification.
• A clinical trial data analysis illustrates the practical application and effectiveness of the developed method.

Conclusions:

• The enhanced doubly robust estimator with concave link functions offers a valuable advancement in causal inference.
• This method provides more reliable and unbiased estimates of treatment effects in challenging observational and quasi-experimental settings.
• The findings contribute to more accurate causal effect estimation in biostatistics and related fields.