Non-REM sleep EEG spectral analysis in insomnia

There is a need for objective measures of insomnia that can be used in diagnosis and outcome assessment. Although PSG is a valuable outcome assessment tool in a selected subset of insomnia patients, we lack measures with diagnostic utility and that can reflect outcome in the entire insomnia population. Non-REM EEG spectral measures have the potential to provide improved measures of the nature of the sleep attained. They reflect the non-REM EEG signal frequency content that is extracted primarily via an algorithm referred to as the fast Fourier transform. The frequency content of the EEG is reflected in the amplitude of a series of frequency bands. Although sleep stages provide a five-level categorical characterization of sleep, spectral indices are continuous measures allowing a more fine-grained characterization. Also, analysis of the pattern of spectral indices over time can allow a dynamic analysis of sleep physiology. The available studies employing spectral analysis in insomnia suggest that there may be greater high frequency and diminished low-frequency EEG activity in at least a subset of insomnia patients. There is some data suggesting that this is the case for those with insomnia who tend to underestimate their sleep, compared with traditional PSG indices of sleep quantity.10,20 This suggests the possibility that spectral indices may reflect different aspects of sleep than traditional indices and, therefore, may be complementary measures with the potential to be combined. Notably, there are few data on the use of spectral indices as treatment outcome measures. Further studies employing spectral indices to assess outcome are needed, particularly where: 1) analyses of the relationship of improvement in self-reported sleep and spectral indices are carried out, and 2) where subjects are recruited who don't meet traditional PSG entry criteria or who have alterations in non-REM EEG frequency content. In addition, there is a need to standardize methods of spectral analysis. It will also be important to determine how/whether to combine spectral indices with traditional PSG or other measures to optimize diagnosis and outcome assessment.

Duke Scholars

Author Andrew Darrell Krystal Psychiatry & Behavioral Sciences, Behavioral Medicine & Neur ...