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

With the increasing popularity of using electroencephalography (EEG) to reveal the treatment effect in drug development clinical trials, the vast volume and complex nature of EEG data compose an intriguing, but challenging, topic. In this paper the statistical analysis methods recommended by the EEG community, along with methods frequently used in the published literature, are first reviewed. A straightforward adjustment of the existing methods to handle multichannel EEG data is then introduced. In addition, based on the spatial smoothness property of EEG data, a new category of statistical methods is proposed. The new methods use a linear combination of low-degree spherical harmonic (SPHARM) basis functions to represent a spatially smoothed version of the EEG data on the scalp, which is close to a sphere in shape. In total, seven statistical methods, including both the existing and the newly proposed methods, are applied to two clinical datasets to compare their power to detect a drug effect. Contrary to the EEG community's recommendation, our results suggest that (1) the nonparametric method does not outperform its parametric counterpart; and (2) including baseline data in the analysis does not always improve the statistical power. In addition, our results recommend that (3) simple paired statistical tests should be avoided due to their poor power; and (4) the proposed spatially smoothed methods perform better than their unsmoothed versions.

[1]  Ernst Fernando Lopes Da Silva Niedermeyer,et al.  Electroencephalography, basic principles, clinical applications, and related fields , 1982 .

[2]  E. John,et al.  Neurometrics: computer-assisted differential diagnosis of brain dysfunctions. , 1988, Science.

[3]  F. H. Lopes da Silva,et al.  Computer-assisted EEG diagnosis: pattern recognition and brain mapping , 1998 .

[4]  William N. Venables,et al.  Modern Applied Statistics with S , 2010 .

[5]  Mike Rees,et al.  5. Statistics for Spatial Data , 1993 .

[6]  Devan V Mehrotra,et al.  Should baseline be a covariate or dependent variable in analyses of change from baseline in clinical trials? , 2009, Statistics in medicine.

[7]  Noel A Cressie,et al.  Statistics for Spatial Data. , 1992 .

[8]  F. H. Lopes da Silva,et al.  Biophysical aspects of EEG and magnetoencephalogram generation , 1998 .

[9]  M R Crager,et al.  Analysis of covariance in parallel-group clinical trials with pretreatment baselines. , 1987, Biometrics.

[10]  F. L. D. Silva,et al.  EEG analysis: Theory and Practice , 1998 .

[11]  EEG Global Field Power Spectrum Changes After a Single Dose of Atypical Antipsychotics in Healthy Volunteers , 2003, Brain Topography.

[12]  D. Bates,et al.  Mixed-Effects Models in S and S-PLUS , 2001 .

[13]  Thomas Dierks,et al.  A pharmaco-EEG study on antipsychotic drugs in healthy volunteers , 2007, Psychopharmacology.

[14]  R. Luthringer,et al.  IPEG Guideline on Statistical Design and Analysis for Pharmacodynamic Trials , 1999, Neuropsychobiology.