Quantile Effect on Duration of Response: A Zero-Inflated Censored Regression Approach.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Survival Analysis

Background:

• Duration of response (DOR) is a critical endpoint in randomized clinical trials (RCTs).
• Traditional DOR estimands like restricted mean DOR can be insensitive to outliers and quantile effects.
• Existing quantile regression methods are not directly applicable to DOR data due to non-responders.

Purpose of the Study:

• To develop a flexible statistical approach for analyzing DOR in RCTs, specifically addressing the challenge of non-responders.
• To accurately model treatment effects on various quantiles of DOR, such as the proportion of patients with a DOR of at least 3 months.
• To provide a robust method for evaluating treatment efficacy beyond simple mean measures.

Main Methods:

• Proposed a novel two-part, zero-inflated quantile regression model tailored for DOR data.
• Modeled non-responders as a distinct component of the model.
• Applied quantile regression to model DOR for responders.

Main Results:

• A simulation study demonstrated the performance of the proposed approach.
• The method was illustrated using a simulated acute myeloid leukemia trial dataset.
• Asymptotic properties of the new statistical approach were derived.

Conclusions:

• The proposed two-part, zero-inflated quantile regression model offers a flexible and accurate method for analyzing duration of response (DOR) in clinical trials.
• This approach effectively handles non-responders, providing more reliable estimates of treatment effects on DOR quantiles.
• The method enhances the evaluation of treatment effectiveness in settings with non-ideal response patterns.