Cumulative incidence regression for dynamic treatment regimens

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Personalized Medicine

Background:

• Dynamic treatment regimens (DTRs) are crucial for personalized medicine.
• Sequential Multiple Assignment Randomized Trials (SMARTs) are key for DTR data collection.
• Existing DTR regression methods do not address competing risks using cumulative incidence functions (CIFs) in SMARTs.

Purpose of the Study:

• To develop and validate novel regression models for DTRs in two-stage SMARTs with competing risks.
• To extend existing CIF regression models to incorporate covariate effects for DTRs within SMART designs.
• To provide a method for analyzing complex DTR data in clinical trials.

Main Methods:

• Extension of cumulative incidence function (CIF) regression models to accommodate DTRs in SMART data.
• Development of estimators with established asymptotic properties.
• Utilizing an augmented-data approximation for practical implementation in existing software.

Main Results:

• The proposed estimators demonstrate asymptotic properties.
• The methods are implementable using standard statistical software.
• Simulations confirm the improvement offered by the new methods over existing approaches.
• Practical utility is shown through a neuroblastoma SMART study analysis.

Conclusions:

• The developed regression models provide a robust framework for analyzing DTRs in SMARTs with competing risks.
• This research fills a critical gap in the statistical methodology for DTR analysis in clinical trials.
• The approach enhances the ability to make informed decisions in personalized medicine based on complex trial data.