Assessing Statistical Methods for Estimating the Cerebrospinal Fluid Early Fungicidal Activity as an Endpoint for...

Area of Science:

• Infectious Diseases
• Clinical Trials
• Statistical Modeling

Background:

• Cryptococcal meningitis is a critical central nervous system infection.
• Early fungicidal activity (EFA) is a key endpoint in phase 2 antifungal trials for cryptococcal meningitis.
• EFA measures the rate of *Cryptococcus* yeast clearance from cerebrospinal fluid (CSF).

Purpose of the Study:

• To systematically compare the 2-step linear regression and linear mixed model approaches for estimating EFA in cryptococcal meningitis.
• To evaluate the performance and operating characteristics of these statistical methods.
• To provide guidance on appropriate statistical methods for analyzing EFA data.

Main Methods:

• Conducted a systematic literature review of EFA estimation methods in cryptococcal meningitis trials.
• Performed simulation studies to assess method performance under various scenarios.
• Applied both the 2-step approach and linear mixed models to phase 2 trial data.

Main Results:

• The 2-step and mixed model approaches yielded discrepant EFA estimates.
• The 2-step approach showed steeper estimates and larger treatment effects.
• Linear mixed models produced smaller standard errors but were biased toward the null with early culture sterility.

Conclusions:

• The 2-step approach is recommended for EFA estimation in cryptococcal meningitis studies, particularly when early culture sterility is observed.
• EFA estimates and comparisons across studies must account for the statistical models and data preprocessing used.
• Clarification of statistical methods is crucial for accurate interpretation of EFA in clinical trials.