Treatment selections using risk-benefit profiles based on data from comparative randomized clinical trials with...

Area of Science:

• Biostatistics
• Clinical Trials
• Personalized Medicine

Background:

• Standard clinical trial analyses often overlook complete patient outcome data.
• Competing risks and censoring limit traditional time-to-event analyses.
• A comprehensive risk-benefit profile requires utilizing all observed event times.

Purpose of the Study:

• To develop a novel statistical framework for analyzing multiple, potentially censored, event times in clinical trials.
• To create a method for classifying patient risk-benefit profiles into clinically meaningful ordinal categories.
• To establish a procedure for identifying patient subgroups who benefit from specific treatments.

Main Methods:

• Classification of patient risk-benefit profiles using all event times.
• Inference methods for censored ordinal categorical data in a two-sample setting.
• Cross-validation for model building and evaluation in personalized medicine.
• Application to a clinical trial for beta-blocker treatment in heart failure.

Main Results:

• Demonstrated a method to handle incomplete outcome data by considering all events.
• Developed a systematic procedure for personalized treatment benefit identification.
• Validated the approach using cross-validation and an independent dataset.

Conclusions:

• The proposed method offers a more complete assessment of treatment effects than standard analyses.
• This framework supports personalized medicine by identifying individual patient benefits.
• The approach is applicable to real-world clinical trial data, such as heart failure treatment.