Statistical methods to estimate treatment effects from multichannel electroencephalography (EEG) data in clinical...

Area of Science:

• Neuroscience
• Biostatistics
• Clinical Trials

Background:

• Electroencephalography (EEG) is increasingly used in clinical trials for drug development.
• Analyzing the large and complex nature of EEG data presents significant statistical challenges.

Purpose of the Study:

• To review and compare statistical methods for analyzing multichannel EEG data in clinical trials.
• To propose and evaluate novel spatially smoothed statistical methods for EEG data analysis.

Main Methods:

• Reviewed existing statistical methods recommended by the EEG community and used in literature.
• Introduced an adjustment for multichannel EEG data.
• Proposed new methods based on spatial smoothness using spherical harmonic (SPHARM) basis functions.
• Applied seven statistical methods to two clinical datasets.

Main Results:

• Nonparametric methods did not outperform parametric counterparts.
• Including baseline data did not consistently improve statistical power.
• Simple paired statistical tests demonstrated poor statistical power.
• Proposed spatially smoothed methods outperformed their unsmoothed versions.

Conclusions:

• Spatially smoothed statistical methods offer improved power for detecting drug effects in EEG clinical trials.
• Careful selection of statistical methods is crucial for effective EEG data analysis in drug development.
• Findings challenge some conventional recommendations within the EEG community regarding statistical analysis.