Analysis of a Binary Outcome Dichotomized from an Underlying Continuous Variable in Clinical Trials

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Statistical Modeling

Background:

• Binary outcomes are common in clinical trials, often derived from continuous variables (e.g., HbA1c in diabetes trials).
• Traditional methods directly estimate proportions from the binary variable.
• An alternative involves using the underlying continuous variable's density for estimation.

Purpose of the Study:

• To compare density estimation methods with direct proportion estimation for binary endpoints derived from continuous data.
• To evaluate the performance of these methods, particularly in early-phase clinical trials.

Main Methods:

• Proposed several density estimation techniques for binary endpoints derived from continuous variables.
• Conducted extensive simulation studies using real clinical trial data.
• Compared the mean squared error (MSE) of density estimation approaches against direct proportion estimation.

Main Results:

• Density estimation methods generally yielded a smaller mean squared error compared to direct estimation.
• This advantage was particularly notable in early-phase studies with limited sample sizes.
• Density estimation introduces bias but offers a potentially favorable bias-variance trade-off.

Conclusions:

• Density estimation approaches are attractive for analyzing binary endpoints derived from continuous variables in early-phase clinical trials.
• These methods can provide more precise estimates when sample sizes are small, despite potential bias.
• The favorable bias-variance trade-off supports the use of density estimation in specific clinical trial contexts.