Review

The use of the EEG in measuring therapeutic

drug action: focus on depression and
antidepressants Journal of Psychopharmacology
25(9) 1175–1191
! The Author(s) 2011
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0269881110388323
Hamid Alhaj1,2, Grzegorz Wisniewski3 and jop.sagepub.com
R Hamish McAllister-Williams1,2

Abstract
A major issue in proof of concept studies and early clinical trials of novel therapeutic agents is that the active drugs can often have a relatively small
additional effect compared with placebo. This is especially the case in psychiatry when we usually have no direct method of measuring the pathology
underlying the disorder being studied but, rather, have to rely on the subjective assessment of psychiatric symptoms. The use of the electroenceph-
alogram (EEG) offers two potential major means of addressing this problem. First it is able to provide direct data relating to neural activity that may be
abnormal in certain disorders. As such there are opportunities for utilizing the EEG in a variety of ways as an objective outcome measure. Second there is
growing evidence that in certain circumstances the EEG can be used to predict which patients are likely to respond to treatment, thus potentially
increasing the power of studies by decreasing non-response rates and increasing mean changes in outcome measure. Both of these uses of the EEG are
illustrated in reference to the study of mood disorders and in particular depression and its treatment with antidepressants.

Keywords
Antidepressants, depression, EEG, ERP, HPA axis

Introduction
cross-over or longitudinal studies, which can be particularly
For many working in the area of psychopharmacology, expo- useful in psychopharmacological investigations. This is not to
sure to electroencephalographic (EEG) methodology will be say that EEG is not without important limitations. Most
extremely limited. With the exception of screening com- notably, the EEG has much lower spatial resolution com-
pounds in an industry setting, many pre-clinic scientists work- pared with fMRI or PET and is extremely limited in its ability
ing with animals will see EEG techniques as being redundant to detect changes occurring deep in the brain. Central to this
compared with single unit or single cell electrophysiological problem is that localizing particular potential changes at the
recordings. Clinical psychiatrists may see the EEG as being scalp to an electrical source within the three-dimensional
important, though not essential, to the diagnosis of epilepsy space of the brain does not have a unique solution. Various
and otherwise its use being limited to such things as the diag- electrical source localization techniques have been developed
nosis of lithium induced neurotoxity. Outside of epileptology, that attempt to cope with this problem but all, necessarily,
sleep research and experimental cognitive neuropsychology, have to be based on one or more assumptions to derive a
EEG methods often receive little attention. However, EEG solution. This lack of spatial resolution means that while
and event-related potential (ERP) techniques can be seen as the EEG can tell us much about ‘when’, and in what order,
an additional form of neuroimaging technique that offers processes are occurring within the brain, it is usually better to
complementary strengths to other methods such as functional use other complementary techniques such as fMRI to identify
magnetic resonance imaging (fMRI) and positron emission ‘where’ they are located. However, there are limitations to
tomography (PET). Importantly, EEG directly reflects neu- this complementary use of EEG and fMRI. The EEG may
ronal electrical activity, rather than being a proxy, as is the
case with fMRI, which reflects blood oxygen concentration 1
Institute of Neuroscience, Newcastle University, Newcastle, UK.
changes. Further, certain EEG methods (as well as the related 2
Northumberland Tyne and Wear NHS Foundation Trust, Newcastle, UK.
technique of magneto-encephalography, (MEG)) have greater 3
Clare Mental Health Services, Co Clare, Ireland.
temporal resolution compared with other neuroimaging tech-
niques. EEG does not involve potentially invasive procedures, Corresponding author:
such as high field-strength magnets or radioactive materials, it Hamish McAllister-Williams, Academic Psychiatry, Newcastle General
is easy to administer, well tolerated and is less expensive to Hospital, Newcastle upon Tyne NE4 6BE, UK
use. This facilitates multiple recordings in subjects, such as in Email: [email protected]
1176 Journal of Psychopharmacology 25(9)

demonstrate multiple neuronal generators of the electrical directed to a number of sources. The textbook of Ernst
activity being recorded temporally separated on a millisecond Niedermeyer and Lopes Da Silva is an extremely comprehen-
time scale. It is often not currently possible to differentiate sive review of the basis of the EEG and its use in a variety of
these sources of activity with fMRI given the time scale of the situations, with a strong clinical focus (Niedermeyer and
BOLD signal being three to four orders of magnitude slower. Da Silva, 2004). It contains relatively little information per-
This is intentionally a selective review, with the goal of taining to the psychiatric issues raised in this review and so
raising awareness of the potential for the use of EEG and further references are cited throughout the text. An excellent
ERP techniques in experimental medicine. In particular it is source for discussion of some of the methodological issues,
intended to raise awareness of the potential for these tech- limitations and fallacies in the use of EEG is the textbook by
niques to be used to address issues of ‘signal to noise ratio’ in Nunez and Srinivasan (Nunez and Srinivasan, 2006). In terms
proof of concept studies and early clinical trials of novel ther- of the analysis of EEG and ERP recordings, and their publi-
apeutic agents. In referring to ‘signal to noise ratio’ we are not cation, a long standing reference source is a consensus paper
meaning this in the specific sense it is used in EEG method- published by a number of authors in 2000 (Picton et al.,
ology (separating out the electric signal of interest from other 2000). This can be usefully supplemented with the multi-
brain and artefact electrical activity). Rather, we are referring authored book edited by Handy (Handy, 2004).
to the major problem of detecting a ‘signal’ of effect of a
psychotropic agent, especially in patient populations. This is Understanding the EEG
reflected by the relatively small drug–placebo difference in
clinical trials that has led some to argue that, for example, A full understanding of the physiology and physics of EEG
antidepressants are ineffective (Kirsch et al., 2008). The small recordings is outside the scope of this review. The reader is
‘signal’ of effect, relative to the ‘noise’ in the measurement of referred to the general references around EEG methodology
psychiatric symptoms and the effect of placebo response may cited in the Introduction. There are many issues and limita-
in part result from the subjectivity of reporting the severity of tions in the recording and interpretation of an EEG. A major
psychiatric symptoms and the heterogeneity of the patients issue in this regard is the general lack of consensus and con-
within a diagnostic category (McAllister-Williams, 2008). sistency in many of the technical issues involved. The EEG
EEG methods, along with other complementary neuroimag- signal is an electrical potential. As such this has to be mea-
ing techniques, offer the possibility to more precisely define sured between two points. One of these will (usually) be an
the pathophysiology of major psychiatric illnesses, predict electrode on the scalp overlying the cortex. The other, the
which patients may be likely to respond to a specific treatment ‘reference’ electrode, can be placed in one of many different
and provide an objective outcome measure of the effect of the locations, for example the nose or the ear lobes, for which
treatment. As such they have the potential to decrease some there is no over-riding consensus. The issue is that all refer-
of the ‘noise’ inherent in the experimental evaluation of puta- ence points, wherever they are located, are not neutral or
tive psychotropics. An important caveat to this contention is inactive since voltage potentials are conducted throughout
that EEG and other imaging techniques in reflecting physio- the body. Because the potential at the different reference loca-
logical processes do not necessarily offer outcome measures tions can be different, where the reference electrode(s) is/are
that are either more valid or more reliable than the qualifica- located can have a major impact on the EEG signal recorded
tion of subjective symptoms. However, rating psychiatric and hampers comparisons between EEG studies. This issue,
symptoms is inherently complex and contentious depending and the problems that arise from it, is discussed in the book
on arbitrary conceptualizations of the disorder under review by Nunez and Srinivasan (Nunez and Srinivasan, 2006).
and the problem of comparing the impact of severity between A number of potential solutions to this issue have been pro-
individuals. This is graphically illustrated by the vast number posed including the computation of a virtual reference point
of symptoms rating scales and the common practice of utiliz- that approaches infinity and hence is subject to minimal activ-
ing more than one in any given study (e.g. Hamilton ity (Yao, 2001).
Depression Rating Scale, Montgomery Asberg Depression Accepting the limitations with regard to reference points
Rating Scale and Clinical Global Impression in studies of that are not inactive, EEG records potentials that directly
antidepressants). result from electrical activity of neurones. This is generated
The current review will focus on the utility of EEG and by inhibitory and excitatory postsynaptic potentials at corti-
ERP studies in the investigation of depression and its treat- cal neuronal synapses, although the polarity of the EEG does
ment with antidepressants through discussion of a number of not reflect whether this is a summation of activation or inhi-
pertinent examples from the authors’ own laboratories as well bition in a certain neuronal network. For an electrical signal
as data from other groups. We have confined ourselves to to be detectable at the scalp a collection of neurones must be
awake EEG data and do not discuss the potential use of firing simultaneously in a synchronous fashion with pyrami-
sleep EEG analysis. The reader is directed to similar reviews dal cells orientated in parallel at 90 to the scalp surface. This
related to schizophrenia (Korostenskaja and Kahkonen, is an important limitation of EEG techniques. Non-synchro-
2009; Van der Stelt and Belger, 2007), the pharmacology of nous neuronal activity that will not be present within an EEG
benzodiazepines (Valle et al., 2002) and pharmaco-EEG more signal may, however, be detectable using fMRI techniques.
generally (Carozzo et al., 2006; John and Prichep, 2006; Conversely, neuronal activity that is not associated with
Mucci et al., 2006) for a wider perspective. This paper is a BOLD oxygen signal and hence not detectable by
not intended as a comprehensive review of all the methodo- fMRI may lead to an EEG signal. This emphasizes the
logical issues around the use of EEG. For this the reader is complementary nature of different neuroimaging techniques.
Alhaj et al. 1177

An additional limitation to the EEG is that it mainly records

electrical activity from superficial cortical areas but even this 0.02
Delta Theta Alpha Beta
signal is attenuated and modified by the tissues it passes
through (CSF, skull and scalp). The degree of attenuation is

Relative power
determined by the degree of synchronization of the postsyn-
aptic potentials, the orientation of the neurones and the size
of the participating area of cortex. All of these factors militate 0.01
against localizing the source of the electrical activity recorded
at the scalp. Therefore, it is difficult to identify the neural
generators within the brain from signals recorded on the
scalp, a difficulty that is called the EEG ‘inverse problem’,
because there are an infinite number of theoretical electrical 0.00
generators that can produce identical potentials at the scalp 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
surface (Baillet et al., 2001; Michel et al., 2004). Finding
Frequency (Hz)
a solution to the inverse problem, the three-dimensional
localization of the sources of voltage fields detected on
the scalp, has been a focus for many researchers over the Figure 1. Fast Fourier Transformation (FFT) of resting EEG, eyes
last two decades. A unique solution to the localization of open, recorded from 14 healthy male subjects. The graph illustrates
scalp electrical activity requires additional constraints to relative power plotted against frequency. Relative power has been
be placed on the data, such as assuming that the poten- calculated by dividing the absolute power at any given frequency
tials are generated by a specified small number of dipolar by the total power between 0 and 30 Hz for each individual subject.
sources, which is used in the ‘equivalent current dipole’ This decreases inter-subject variation due to total spectral power
techniques (Koles, 1998; Scherg and Von Cramon, 1986). differences, for example due to differences in signal transduction
Other methods have been developed making different through the individual’s skull, scalp and EEG electrodes. Delta,
assumptions. including constraining the solution to areas theta, alpha and beta frequency bands are marked. Note that there
of interest within grey matter, including brain electrical is considerable variability in the exact ranges of these bands between
source analysis, weighted minimum norm estimation and research groups. (Data adapted from McAllister-Williams and Massey,
Low-Resolution Electromagnetic Tomographic Analysis 2003).
(LORETA) (Pascual-Marqui et al., 1994), or using anatomi-
cal and physiological information derived from other imaging
modalities (Phillips et al., 2002).
In the past, EEG was recorded from a few scalp electrodes, ERP waveforms may consist of a number of distinctive
often placed on the midline of the head, with an analogue peaks and troughs (positive and negative fluctuations of
signal passing to a pen and paper chart recorder. The resul- voltage – see Figure 2). These are sometimes rather simplis-
tant traces were analysed simply by inspection to identify tically referred to as ‘components’ with an implication that
aberrant activity, particularly which associated with epilepsy. they relate to different and distinct underlying neurocogni-
Frequencies seen in the EEG were divided into four main tive processes. Various components have been identified and
groups, namely delta < 4 Hz; theta 4–8 Hz; alpha 9–13 Hz; named according to the approximate times (in milliseconds)
and beta > 13 Hz. Nowadays EEG recordings can be made at which they occur post-stimulus and the direction of the
from tens, if not hundreds, of electrodes placed across the deflection of the waveform (N: negative vs. P: positive). For
scalp (though this does not in itself assist with the inverse example, it is believed that N100 and P200 are directly
problem (Ollikainen et al., 2000)). Signals are digitized, allow- evoked by a sensory stimulus (e.g. visual vs. auditory).
ing computer based analysis, such as Fast Fourier In contrast, components such as N200 and P300 are thought
Transformations (FFTs) that calculate the ‘amount’ of differ- to be associated with the psychological processes invoked by
ent frequencies in the whole EEG, and attempts at source the stimulus. However, the notion that a particular peak or
localization (see Figure 1). trough reflects a single component of brain activity is a
major oversimplification. Activity related to multiple brain
Event-related potentials sources may be superimposed to produce a particular wave-
form. A voltage measured at particular time point and a
Event-related potentials (ERPs) are alterations in the voltage particular scalp location may represent the activity of mul-
of the EEG due to various neurocognitive, sensory and motor tiple ERP components. Further, the location of the reference
activities. These ERPs are relatively small voltage changes electrode can profoundly affect the shape and time course of
overlaid on background EEG activity that is unrelated to the ERP waveform confounding simple peak detection as
the process or event of interest. To extract the ERP of inter- component identification. Rather, a component is defined
est, EEG is recorded time-locked to the presentation of the by having a specific topography and time course, and is
stimuli or responses of interest and averaged together to related in a characteristic way to the experimental manipu-
increase the signal-to-noise ratio (Figure 2). The reader is lations being conducted (Picton et al., 2000). These compo-
referred to other sources for more detailed description of nents can be separated out, for example, using either a
ERP recording methodology and analysis (Handy, 2004; spatial or temporal Principal Component Analysis (PCA)
Nunez and Srinivasan, 2006; Picton et al., 2000). (Picton et al., 2000).
1178 Journal of Psychopharmacology 25(9)

(A) Stimulus 1 Stimulus 2 Stimulus N

EEG

(B)
Averaged ERP
Stimulus 1 –
N1
N2

Voltage (mV)
Stimulus 2
P1 P2

P3
Stimulus N +
0 200 400 600
Time (ms)

Figure 2. Illustration of the principle behind ERP analysis and some commonly identified peaks in the ERP waveform. (A) Stimuli (1. . .N) are presented
while the EEG is recorded. The resultant electrical activity is relatively small in comparison to the background EEG and is hence not visible. (B) However,
averaging the EEG segments or ‘epochs’ time locked to the presentation of the stimulus leads to the ERP components becoming identifiable due to the
non-time locked EEG activity being averaged out. The ERP illustrated is what may be obtained in an auditory ‘oddball’ experiment (see section The P300
in depression). Note that by convention EEG and ERP data are plotted with negative deflections going upwards and positive downwards. Also note that
positive and negative peaks are referred to (and labelled here) as P and N respectively. The first positive peak is further described as P1, the second P2
etc. Peaks are also described by the time in which they are seen, so the P3 is also referred to as P300 since it is often seen at around 300 ms post stimuli
(though with considerable variable and in the example shown – 400 ms). Peaks in the ERP waveform should not be seen as simple ‘components’ since
they may reflect multiple electrical sources within the brain (see text in section Event-related potentials).

components that have been postulated to represent endophe-

EEG/ERPs as measures of abnormalities in notype measure of depression due to their heritability and
depression stability over time (Smit et al., 2005). It is important to
note that there are few suggestions in the literature that
Depressive disorders are associated with changes to the EEG and ERP findings related to depression might be
normal electrophysiology and neurocognitive processing. useful in the diagnosis or detection of the disorder in a similar
Given that EEG is the ideal method to diagnose epileptic way to the use of the EEG to diagnose epilepsy. Rather the
seizures, demonstrating the link between depression and epi- findings are similar to most biological investigations in psy-
lepsy is perhaps an obvious starting point to understanding chiatric conditions – they are relatively small and subtle dif-
EEG changes in depression. Epidemiological studies demon- ferences between patients and matched controls. These
strate that up to half of patients with epilepsy have clinically findings, while not necessarily of direct clinical utility, do
significant depression and/or anxiety (Harden, 2002). offer possibilities with regard to further understanding the
Interestingly, depression is more common in patients with nature and underlying pathophysiology of depression.
left temporal lobe epilepsy (Altshuler et al., 1990), suggesting
that it is not simply psychosocial factors that account for the
EEG frequency band changes in depression
high rates of comorbidity. The nature and even the direction
of the causality between depression and epilepsy is not clear FFT analysis is used to calculate signal power against fre-
given findings that depressive illness is associated with a six- quency and the resulting data are usually described in relation
fold increase in the risk of the development of seizures in an to certain frequency bands. High frequency EEG activity such
elderly population (Hesdorffer et al., 2000). It is possible that as beta (13–30 Hz) and gamma (>30 Hz) appears to play an
the association between the two conditions reflects some important role in various cognitive processes including atten-
common underlying pathology and it has been argued that tion (Fell et al., 2002; Wrobel, 2000). Gamma activity may
depression may simply be a behavioural form of seizure dis- also play a role in memory, including ‘binding’ together the
turbance (Zyss, 2009). This section will attempt to address the features of a perceived object and comparing this with repre-
EEG alterations seen in depression, including changes in sentations in memory (Herrmann et al., 2010). In contrast,
power spectral density, hemispheric asymmetry and ERP low frequency EEG activity including theta (4–8 Hz) and
Alhaj et al. 1179

delta (<4 Hz) are seen in slow-wave sleep (Kubicki and A further observation that may be important in a full
Herrmann, 1996), though theta has also been shown to play understanding of the pathophysiology of depression is that
a role interacting with alpha activity (approximately 8–13 Hz) of a temporal coupling between EEG power in the high fre-
in cognitive tasks such as the transfer of information between quency activity (beta and gamma frequencies) and secretory
memory systems (Sauseng et al., 2002). Interestingly, EEG- pulses of cortisol (Chapototet al., 2001; Gronfier et al, 1998).
neurofeedback aimed at increasing the ratio of theta to alpha This is interesting since it may provide an explanation to the
has been shown to improve creative performance and mood variations in arousal seen in depression, known to be linked
(Gruzelier, 2009). This emphasizes the important interplay with altered function of the hypothalamic–pituitary–adrenal
between the brain functions underlying the various frequency (HPA) axis.
bands as well as the fact that like ERP voltage peaks, it is an
incorrect oversimplification to associate an EEG frequency
EEG asymmetry in depression
band with a single cognitive function.
Changes in the normal frequencies of EEG have been EEG asymmetry of cortical activity reflects a general concept
demonstrated in depressed patients, although with some of brain behaviour that may occur in states of emotional dis-
inconsistent results (Coburn et al., 2006; Guidi et al., 1989; turbance and stress (Alhaj et al., 2008). In particular, there is
Kwon et al., 1996). In their review, Pollock and Schneider evidence suggesting that depression is characterized by func-
(1990) described increased alpha and beta power in over a tional asymmetric cortical activity (Allen et al., 2004; Coan
half of un-medicated depressed patients (Pollock and and Allen, 2004; Davidson, 1998; Henriques and Davidson,
Schneider, 1990). Similarly, increases in overall relative and 1991). For example, Henriques and Davidson (1991) have
absolute beta power has been reported in a study of 70 male found that depressed patients show increased left frontal
drug-free depressed unipolar patients compared with 23 alpha activity compared with healthy subjects. This increase
normal controls (Knott, 2000) and increases in delta and in alpha activity has been frequently shown to correlate with
theta power over the right hemisphere in 20 patients versus functional cortical hypoactivation (Laufs et al., 2003;
matched controls (Kwon et al., 1996). Leuchter et al., 1999). Davidson and Henriques proposed a
LORETA localization of EEG frequency bands may be a hypothesis of frontal alpha asymmetry pattern in depression,
more sensitive method of examining differences in EEG in which left frontal hypoactivation is interpreted as a deficit
power between depressed patients and healthy controls. In a in approach mechanisms, whereas right frontal hypoactiva-
study of 37 drug-free depressed patients and 37 matched con- tion is associated with deficits in the withdrawal system
trols, Korb et al. (2009) found no significant differences in (Davidson and Henriques, 2000). A relationship between
global spectral power. However, LORETA analysis (using a frontal alpha asymmetry and negative emotions has been sup-
high density of frontal EEG electrodes) revealed elevated ported by additional functional neuroimaging methods, such
delta, theta and beta power in anterior cingulated cortex as PET (Holthoff et al., 2004), single photon emission tomog-
(ACC), and increased alpha power in ACC, dorsolateral raphy (SPECT) (Mottaghy et al., 2002) and fMRI (Killgore
prefrontal cortex (DLPFC) and lateral orbito-frontal et al., 2007). However, frontal alpha asymmetry has not
cortex (OFC) (Korb et al., 2008), areas of the brain shown always been found in depression (e.g. Debener et al., 2000;
using other neuroimaging techniques to be functionally Reid et al., 1998) and it has been suggested that this discrep-
abnormal in depressed patients (Frodl et al., 2010; Mayberg ancy may have occurred due to heterogeneity of depressed
et al., 1999). However, there remains a question of what the patient groups and methodological differences used in differ-
changes in EEG frequency bands in depression actually mean. ent labs (Davidson, 1998), in particular, the location of the
Grin-Yatsenko et al. using an Independent Components reference electrode can influence whether or not asymmetry is
Analysis (ICA) of EEG recorded from 111 patients with seen (Hagemann, 2004). Furthermore two meta-analyses of
depression compared with 526 non-depressed subjects found functional imaging studies have failed to show consistent hae-
an increase in theta and alpha activity in the patients at modynamic asymmetries related to emotion (Murphy et al.,
parietal and occipital electrode sites (Grin-Yatsenko et al., 2003; Wager et al., 2003). Nevertheless it is argued that a
2010). They argued that this may reflect decreased cortical frontal EEG asymmetry can be recorded in a reliable way
activation in these brain regions. However, the lack of spatial (Towers and Allen, 2009) and is supported by a meta-analysis
resolution of EEG means that other neuroimaging techniques of the data from the various studies described above
would be preferred in the identification of the cortical areas (Thibodeau et al., 2006). In addition, the inconsistent findings
that are functionally abnormal in depression. Where EEG from haemodynamic imaging studies have been challenged by
frequency analysis may play a role is when there is a strong a study confirming that emotional stimulus processing and
link between the frequency band and cognitive processing. trait depression are associated with asymmetric blood flow
For example a group in the USA have shown that gamma in DLPFC that may have been undetected in some previous
band activity is increased in depressed patients following the imaging studies due to inappropriate analytical techniques
presentation of emotionally negative words (Siegle et al., (Herrington et al., 2010). Attempts have also been made to
2010). They suggest that this may represent depression introduce alternative methods of assessing EEG frontal asym-
being associated with an engagement in sustained elaboration metry (actually excluding alpha power) that have been
following emotional stimuli. Given at least the face validity of claimed to be highly discriminatory between depressed
abnormalities in processing negative emotional stimuli in patients and controls and to highly correlate with depressive
depression, this is clearly an area that warrants further symptom severity (Hinrikus et al., 2009). The meaning of a
research. frontal asymmetry is unclear. In reviewing around 70 EEG
1180 Journal of Psychopharmacology 25(9)

studies of the influence of emotion on frontal asymmetry, improvement in depression with rTMS correlates with a
Coan and Allen concluded that it may be ‘a moderator reduction in left–right asymmetry (Mottaghy et al., 2002).
and/or mediator of emotion’ though unequivocal evidence This is in line with the hypothesis of hypoactivity in the left
in support of this is lacking (Coan and Allen, 2004). The relative to the right prefrontal cortex described above, though
recent localization of the asymmetry to DLPFC not all rTMS studies are consistent in this regard (e.g.
(Herrington et al., 2010) is of interest given the role of this Kimbrell et al., 1999; Padberg et al., 1999).
region in various emotional processes, particularly
maintenance of attention on emotional stimuli
The P300 in depression
(Compton et al., 2003; Engels et al., 2007). In addition, of
relevance to this review, the asymmetry appears to be Perhaps the most studied conventional ERP paradigm in
modifiable by pharmacotherapy with venlafaxine (Davidson depression is the auditory evoked P300 (or ‘P3’). This usually
et al., 2003). involves presentation of auditory tones to subjects, the major-
Posterior and temporoparietal alpha asymmetry, with rel- ity of which are identical with random deviant tones occur-
atively reduced right cortical activation, has also been ring at a low frequency (perhaps 1 in 10 tones). When the
reported in patients with depression by many (e.g. Bruder ERPs related to the standard tones are compared with those
et al., 1997; Kentgen et al., 2000) but not all groups related to deviant ones, the ‘deviant ERPs’ show a large pos-
(Henriques and Davidson, 1991). Posterior asymmetry in itive going deflection occurring around 300 ms following the
depression has also been reported in functional neuroimaging onset of the tone (see the ERP example illustrated in
studies with hypoactivation in the right parietal and posterior Figure 2). These ‘P300’ deflections reflect multiple cognitive
cortex (for review see Drevets and Raichle, 1995). processes including attention and higher auditory processing
There is good evidence that this ‘asymmetry hypothesis of and originate from multiple cortical areas (Mulert et al., 2004;
depression’ is not only state-dependent (i.e. related to Volpe et al., 2007). In depression, there are reports of a delay
depressed mood) but perhaps additionally trait-dependent, in the latency of the P300 component (Bruder et al., 1991)
since it has been shown in depressed patients, even while compared with that seen in healthy subjects (Blackwood
euthymic (Henriques and Davidson, 1990). Further, the et al., 1996; O’Donnell et al., 2004; Souza et al., 1995) and
same model of frontal alpha asymmetry has been reported a decrease in the current density in the right hemisphere on
in very young children of depressed mothers (Jones et al., LORETA analysis (Kawasaki et al., 2004), in line with a right
1997) as well as in adult offspring of depressed parents sided hypoactivity.
(Bruder et al., 2005). As such it may relate to fundamental The precise neurobiological basis of the P300 is unknown
underlying pathophysiologies of the disorder, for example but the finding that its latency correlates with the prolactin
reflecting the hypercortisolaemia often seen in depression response to the 5-HT1A agonist flesinoxan (Hansenne and
(Alhaj et al., 2008). Alternatively, or additionally, asymmetry Ansseau, 1999) does make it of interest with regard to anti-
may reflect certain depressive symptoms, such as rumination depressants given the hypothesized importance of these recep-
and low self-esteem (Putnam and McSweeney, 2008). tors in the therapeutic action of at least some antidepressants
It is important to raise a question as to the pathophysio- (Blier and de Montigny, 1994). However, there is very little
logical and clinical relevance of the purported cortical asym- published data on the relationship of P300 and antidepres-
metrical activity found in depression. As has already been sants beyond findings that its latency normalizes after four
mentioned, there have been suggestions that asymmetry weeks’ treatment (Hetzel et al., 2005) and a ‘normal’ P300
scores correspond to individual differences in approach and amplitude predicts response to ECT (Ancy et al., 1996).
withdrawal tendencies (reviewed by Coan and Allen, 2003,
2004) that may be relevant to a predisposition to depression.
EEG and ERP studies of the HPA axis
More recently, there has also been evidence to suggest that
trait EEG asymmetry may be associated with polymorphisms A potentially important pathological phenomenon associated
of the 5-HT1A receptor gene (Bismark et al., 2010), that this with depression is the well documented HPA axis dysfunction
receptor (and its gene polymorphisms) may be central to the (Carroll et al., 1976; Gibbons and McHugh, 1962; Sternbach
pathophysiology of depression (Savitz et al., 2009) and mech- et al., 1983). This section focuses on how the EEG and ERPs
anism of action of antidepressants (Blier and de Montigny, have been used to study the effects of HPA axis manipula-
1994). The importance of cortical activity asymmetry in tions and how this may relate to depression and its underlying
depression is further supported by findings regarding the ther- pathophysiology.
apeutic effect of repetitive transcranial magnetic stimulation EEG and ERP studies of cortisol administration to
(rTMS) in depressed patients (Burt et al., 2002; Gershon healthy subjects have assisted in elucidating the neurobiolog-
et al., 2003; Holtzheimer et al., 2001). One of the main factors ical changes in depression. For example, our results show that
that appears to determine rTMS efficacy is thought to be the a seven-day course of cortisol (20 mg b.d.) as well as a four-
hemispheric site of magnetic stimulation (left vs. right) over day course (100 mg mane and 60 mg nocte) cause qualitative
the DLPFC (Pascual-Leone et al., 1996). In particular, it is changes in the ERP waveforms associated with episodic
believed that for optimal antidepressant effects, rTMS should memory over frontal scalp regions (Alhaj et al., 2008;
be applied in high-frequency (to enhance activation) over the McAllister-Williams and Rugg, 2002). This can be seen in
left DLPFC or low-frequency (to reduce activation) over the Figure 3, which shows the magnitude of the memory ERP
right DLPFC (Bermpohl et al., 2006; Burt et al., 2002; effect following placebo and cortisol treatment. However,
Gershon et al., 2003) and it has been reported that an acute single dose of cortisol (100 mg) had no effect on
Alhaj et al. 1181

The specific involvement of the right frontal cortex following

Placebo
cortisol administration is perhaps due to a right lateralized
3 increase of the monitoring of products of memory retrieval.
4.5 An ERP study of the electrophysiological correlates of
2.5
4 episodic memory revealed greater activation of left frontal

Voltage (mV)

Voltage (mV)
2
3.5 and parietal lobes between 800 and 1100 ms post stimulus in
1.5 healthy controls compared with depressed patients (Alhaj
3
1 et al., 2007). This study also demonstrated a number of cor-
2.5
0.5 relations between cortisol levels and the laterality of electrical
2 activity during episodic memory retrieval with activity. This is
0
1.5 interesting since it suggests that similar to exogenous cortisol
administration, endogenous cortisol may have specific effects
on the processing of episodic memory and, at least in part,
Cortisol account for the hemispheric activity asymmetry related to
5
depression (Alhaj et al., 2007).
2
4
EEG/ERPs as a physiological measure of the
Voltage (mV)

1
3
0 Voltage (mV) effects of antidepressants
2 EEG studies of the effects of psychotropic medications are
–1 quantitative computations of the modulatory effects of the
1 therapeutic drug on the brain activity. EEG has been shown
–2 to be a useful tool in clinical psychopharmacology. For exam-
500-800 ms 800-1100 ms ple, EEG studies can ascertain whether a certain drug has
central effects, demonstrate the minimal central effective
dose and identify the onset and duration of its effect on the
Figure 3. Spherical spline maps illustrating the scalp topography of the
CNS. The EEG can also be used to determine whether a
episodic memory ERP effect following seven days’ treatment with placebo
certain psychotropic medication shows stimulating or sedat-
(upper row) or cortisol 20 mg twice daily. ERPs were recorded during an
ing effects by measuring the relative delta/theta to alpha/beta
episodic memory paradigm and waveforms when no memory was involved
frequencies (de Visser et al., 2003; Knott, 2000; Saletu et al.,
were subtracted from those associated with accurate memory retrieval to
2002). This section will focus on the specific electrophysiolog-
obtain a memory ERP effect. The magnitude of this is plotted over the
ical effects of antidepressants that may be used as a biological
scalp as if looking down on this with the front of the head uppermost. The
measure to develop novel therapeutic drugs.
figure shows following placebo electrical activity occurring over posterior
temporal regions more so on the left than the right together with a
frontal component that becomes prominent over time. However, after Antidepressant effects on EEG frequency bands
cortisol administration, while the temporal component is unchanged, the
Given the differences seen in certain EEG frequency bands in
frontal component is not present. Full details of the study are provided in
patients with depression when drug free (see above), it is of
McAllister-Williams and Rugg (2002).
interest to examine whether these abnormalities are altered by
antidepressant treatment. A number of studies have demon-
the same ERP components though it did alter ERPs related to strated acute effects of antidepressants on EEG power (seen
error processing (Hsu et al., 2003). Similarly, a single dose of within hours after a single dose) in healthy subjects (for a
cortisol (30 mg) does not alter ERPs in a face recognition review see, for example, Saletu et al., 2002) and in depressed
memory task following long-term (30 min) delay, although patients (Galderisi et al., 1996; Knott et al., 1996; Ulrich
it modulated ERPs associated with short-term (6–18 s) et al., 1988). Effects are also seen after up to two weeks of
memory (Monk and Nelson, 2002). Therefore, it is likely treatment (Siepmann et al., 2002; Wisniewski et al., 2008),
that the effect of cortisol on EEG is of a complex nature though longer-term studies of the effects of antidepressant
and probably task-, dose- and time-dependent. drugs on EEG power are less consistent, often with no signif-
Corticosteroids may play a part in the EEG asymmetry icant effects seen (Bruder et al., 2008; Kwon et al., 1996).
seen in depression. It has been shown that four days of Perhaps a more promising avenue is utilizing the EEG to
administration of 160 mg of the synthetic glucocorticoid pred- examine specific central pharmacological effects of various
nisone to healthy subjects alters the laterality of alpha EEG, pharmacological probes. For example, a number of studies
increasing right frontal activation (Schmidt et al., 1999). have consistently shown that the administration of 5-HT1A
Similar findings have also been shown following acute treat- receptor agonists causes a negative shift of the EEG fre-
ment with cortisol in healthy subjects (Tops et al., 2005). ERP quency spectrum. This has been shown with buspirone
studies have also found similar asymmetrical changes in the administration in healthy humans, due to an increase in
neural correlates of episodic memory, causing a more positive theta and a decrease in alpha activity (Anderer et al., 2000;
deflection over the right than the left hemisphere following Barbanoj et al., 1994; Holland et al., 1994; McAllister-
cortisol administration to healthy subjects (Alhaj et al., 2008). Williams and Massey, 2003; McAllister-Williams et al.,
1182 Journal of Psychopharmacology 25(9)

2007; Murasaki et al., 1989). Although buspirone is a rela- is, it may represent a direct method of assessing a central
tively non-selective drug, and its systemic administration action of antidepressants that may be critical to their very
makes the localization of effects difficult, several pieces of mechanism of action.
evidence indicate that this negative shift may be mediated
by somatodendritic 5-HT1A receptors. Firstly, the EEG fre-
quency effects are mimicked by other more selective 5-HT1A Antidepressant effects on Loudness Dependency of
partial agonists in man (Anderer et al., 2000; Saito et al.,
Auditory Evoked Potentials (LDAEPs)
1993) and is seen in animals following administration of bus-
pirone as well as the highly selective full 5-HT1A receptor Hegerl and Juckel (1993) suggested that a potential method
agonist 8-OH-DPAT (Bogdanov and Bogdanov, 1994). for investigating serotonergic function in man is the exami-
Conversely, while buspirone metabolites have dopamine D2 nation of the loudness dependency of auditory event related
receptor antagonist properties (Pruus et al., 2000; Rijnders potentials (LDAEPs). A characteristic ERP waveform can be
and Slangen, 1993), haloperidol in rats does not cause the seen following an auditory stimulus of a single tone. An early
same effect as a 5-HT1A receptor agonist (Bogdanov and negative deflection, N1, and positive deflection, P2, occur
Bogdanov, 1994) and sulpiride has an opposite effect on between 100 and 200 ms post stimulus and their peak to
theta and alpha power (Dimpfel, 2008). Secondly, source peak amplitude depends partially on the loudness of the stim-
localization of the effect of buspirone using LORETA dem- ulus (Figure 4). Hegerl and Juckel suggest that the slope of a
onstrates a significant increase in theta EEG activity in the plot of N1/P2 amplitude against loudness, the amplitude/
hippocampus as well as neighbouring cortical areas (Anderer stimulus intensity function (ASF-slope), inversely correlates
et al., 2000). Hippocampal theta activity is well known to be with serotonergic activity (Hegerl and Juckel, 1993). The most
under ascending serotonergic control (Vertes, 1982). Animal convincing evidence for this hypothesis comes from animal
studies using local application of 8-OH-DPAT into the raphe studies showing that direct injection of a 5-HT1A agonist into
have indicated that activation of somatodendritic 5-HT1A the dorsal raphe nucleus of cats increased the ASF-slope
receptors causes a decrease in hippocampal theta (Nitz and (Juckel et al., 1999), an effect presumed to occur due to soma-
McNaughton, 1999; Vertes et al., 1994), an effect blocked by todendritic 5-HT1A autoreceptor activation decreasing the
5-HT1A antagonists (Marrosu et al., 1996). Thirdly, a soma- firing rate of raphe neurones and so reducing 5HT release
todendritic location for a 5-HT1A receptor mediated negative in the cortex. However, the evidence supporting the hypoth-
shift in the EEG frequency spectrum in man is also inferred esis in man is more circumstantial. For example the ASF-
from findings of acute administration of selective serotonin slope is reported to negatively correlate with plasma 5-HT
reuptake inhibitors (SSRIs) causing such a shift (Saletu et al., concentrations following fluvoxamine (an SSRI) administra-
1986) and that pindolol mimics, rather than blocks, the effect tion, but only in depressed patients and not controls (see
of buspirone on the awake EEG (McAllister-Williams and Hegerl et al., 2001). However, acute administration of
Massey, 2003). Following acute administration, SSRIs SSRIs may actually decrease overall central serotonergic
increase concentrations of 5-HT in the raphe nuclei leading activity due to increased 5-HT activating somatodendritic 5-
to somatodendritic 5-HT1A activation (Bel and Artigas, HT1A autoreceptors (Hjorth and Auerbach, 1994). Direct
1992); pindolol acts as a 5-HT1A antagonist at postsynaptic manipulation of human central 5-HT using the tryptophan
receptors but a partial agonist at somatodendritic receptors depletion paradigm (Moore et al., 2000; Reilly et al., 1997)
(Clifford et al., 1998). Taken together, these data support the appears to have no effect on either N1/P2 amplitude (Phillips
hypothesis that the effect of buspirone on the EEG frequency et al., 2000) or LDAEP (Debener et al., 2002; Dierks
spectrum is a potential index of somatodendritic 5-HT1A et al., 1999; Massey et al., 2004). A recent review has thus
receptor function. As such this presents the possibility of concluded that the LDAEP is not a good index of central
this being a method for studying the effects of antidepressant 5-HT function (O’Neill et al., 2008). While the LDAEP is
treatments, and other drugs, on somatodendritic 5-HT1A reported to be no different in depressed patients compared
function as well as exploring changes in this receptor in with controls (Linka et al., 2007), O’Neill et al. (2008)
patients with depression (McAllister-Williams et al., 2004). argue that the LDAEP does have significant promise as a
potential biomarker of antidepressant effects, since it may
be able to predict response to antidepressants (see the
Antidepressant effects on alpha hemispheric section Loudness Dependency of Auditory Evoked Potentials
below).
asymmetry
As discussed above, hemispheric asymmetry is a well
Antidepressant effects on theta ‘cordance’
described observation in depression that may relate to effects
of hypercortisolaemia on cortical function. This observation EEG cordance is an analysis procedure developed at the
may be of utility in studying the potential antidepressant University of California in Los Angeles (UCLA). It combines
effect of novel medications. Four days of pre-treatment with information from both absolute and relative power from the
the SSRI citalopram has been shown to attenuate the acute EEG spectrum, as well as information from neighbouring
effects of cortisol administration on both EEG alpha power electrodes for each scalp electrode (Leuchter et al., 1994).
and cognitive function in healthy volunteers (McAllister- The analysis algorithm produces two measures – a categorical
Williams et al., 2009). It remains to be seen if this is an value (‘concordant’ versus ‘discordant’) and a numerical
effect common to different classes of antidepressant, but if it value. It has been reported to have a stronger correlation
Alhaj et al. 1183

of fluoxetine (n ¼ 24 patients) and venlafaxine (n ¼ 27

(A)
94 dB patients). In general the group have focused on theta cor-
dance at frontal electrode sites, on the basis that this may
54 dB reflect theta activity in anterior cingulated cortex (Cook
et al., 2002). They have examined the change in cordance
+ from pre-treatment to 48 hours, one or two weeks post-treat-
ment. A decrease in frontal theta cordance at 48 hours (a
trend) and at one week (significantly) predicts clinical
54 dB response to antidepressants with a sensitivity of 69% and
10 mV specificity of 75% (Cook et al., 2002). There is no difference
94 dB between response to fluoxetine or venlafaxine and the final
HDRS score correlates with cordance at week 1 (Cook et al.,
2002). Placebo responders are reported to show a significant
–100 0 100 200 300 400 500 increase in prefrontal theta cordance, while both medication
Time (ms) and placebo non-responders show no change (Leuchter et al.,
2002). Similar findings to those for antidepressants have been
shown by the UCLA group for electroconvulsive therapy
N1 P2 (ECT) treatment (Stubbeman et al., 2004) and by a study
from a Czech group regarding response to buproprion in 18
(B) 18
hospitalized depressed patients (Bares et al., 2010). In treat-
16 ment refractory patients, in a pilot study in 12 depressed
N1/P2 amplitude (mV)

patients who had failed at least six weeks of SSRI monother-

14 apy, five out of six responders, compared with two out of six
12 non-responders to augmentation of switches of treatment,
showed a decrease in prefrontal theta cordance (Cook et al.,
10 2005). This has been replicated by the Czech group firstly in
17 in-patients with treatment resistant depression in which all
8
five responders showed a decrease frontal theta cordance over
6 one week, compared with just two out of 12 non-responders
(Bares et al., 2007). In the most recent study from this group,
4
again in treatment resistant patients (n ¼ 25) a similar finding
50 60 70 80 90 100
predicted response to four weeks’ treatment with venlafaxine
Loundness (dB) (Bares et al., 2008). It remains unclear exactly what the mea-
sure actually reflects, and the findings are difficult to replicate
Figure 4. (A) Loudness Dependency of Auditory Evoked Potentials due to the cordance algorithm being patented. As a result, the
(LDAEPs). Auditory evoked potentials recorded from 14 healthy use of cordance as a biomarker of antidepressant response
male volunteers associated with five different loudness intensities has been questioned (Kuo and Tsai, 2010).
recorded at the Cz electrode. The smallest (solid line) were recorded
associated with 54 dB tones, and largest (also solid lines) with 94 dB
tones, as indicated. The intermediate sized potentials were recorded EEG as a predictor of response to
associated with 64 dB (dotted lines), 74 dB (dashed lines) and 84 dB antidepressant drugs
(dot–dash lines) tones respectively. (B) Auditory evoked potential
amplitude stimulus function. The peak-to-peak amplitudes between the The goal of experimental medicine in relation to the pharma-
N1 and P2 components (means  standard errors) is plotted against cological treatment of mental illnesses can be summarized as
loudness of the auditory stimuli. A linear regression line has been fitted to increase the ‘signal to noise ratio’ in studies investigating
to the data using GraphPad Prism version 3.02. (Data adapted from putative therapeutic agents. Most conventional approaches,
Massey et al., 2004.). including those detailed above in this review, are aimed at
increasing the ‘signal strength’ by using objective precise mea-
sures (e.g. EEG or emotional processing changes in response
to antidepressant treatment) rather than less precise and sub-
with cerebral perfusion than standard spectral analysis jective measures (e.g. changes in symptoms related to the con-
(Leuchter et al., 1999). Further cordance is less influenced cept of depression – although it should be noted that there is
by age, gender and severity of baseline depression that no guarantee that an objective physiological measure is either
simple spectral power (Morgan et al., 2005). The group more reliable or more valid than subjective symptoms).
have mainly confined their research to an analysis of the However, an alternative approach is to ‘decrease noise’ by
theta band. means of identification of individuals who are, and are not,
The UCLA group have published extensively on early likely to respond to a treatment. This is an important consid-
effects of antidepressant treatment on cordance method and eration given that in randomized controlled trials only
its predictive ability in depression. The data that they analyse around 60–70% of depressed patients show a response to a
come mainly from two placebo controlled randomized trials particular antidepressant over an eight-week treatment trial
1184 Journal of Psychopharmacology 25(9)

(Fawcett and Barkin, 1997). Enriching a sample with individ- extent in occipital regions (Suffin and Emory, 1995). Two
uals likely to respond to a given treatment helps to increase other studies that have utilized the neurometrics technique
the power of studies and the chances of signal detection. have used it to investigate antidepressant responses in obses-
To date a number of potential methods of predicting response sive compulsive disorder (OCD). In the first of these, 27
to antidepressants have been explored with lesser or greater patients with OCD were treated with SSRIs. Response was
success. These include pharmacogenetics (for reviews see associated with increased relative power in alpha at baseline,
Perlis, 2007 and Serretti et al., 2005), imaging (Dougherty while non-responders exhibited higher relative theta power,
and Rauch, 2007; Evans et al., 2006; Miller et al., 2008) and especially in frontal and frontotemporal regions (Prichep
neuroendocrine techniques (e.g. Ising et al., 2007). et al., 1993). The reported accuracy rates for determination
This section will consider the promise of a number of EEG of responders and non-responders were 82.4% and 80%
or ERP parameters in predicting antidepressant response, pri- respectively. In the second study in OCD, 20 patients with
marily in relation to unipolar depression. The previous sec- no concurrent depression were treated with paroxetine,
tion discussed the use of early changes in EEG parameters 40–80 mg for at least 12 weeks. Responders were reported
(e.g. cordance) as predictors of clinical response. Here we as having a significant excess of absolute and relative alpha
consider data around the use of ‘baseline’ parameters power especially in the frontopolar, frontal and posterior
recorded prior to treatment with antidepressants. temporal regions (Hansen et al., 2003), essentially replicating
the initial findings. These OCD studies are consistent with
those in depressed patients and hence pre-treatment alpha
EEG alpha power
power may predict the likelihood of antidepressants having
Alpha band is probably the most widely studied in different an effect in general rather than necessarily a specific antide-
theoretical contexts due to its being believed that alpha pressant effect.
reflects cognitive and memory processes (Klimesch, 1999;
Knyazev, 2007), bioelectrical hemispheric asymmetry
Alpha hemispheric asymmetry
(Henriques and Davidson, 1991) and that it is as a reliable
EEG measure of physiological state of vigilance and arousal Given the possible relationship between the hemispheric
(Bruder et al., 2008; Ulrich et al., 1984). Studies of alpha asymmetry described in depression and underlying patho-
power in relation to prediction of antidepressant response physiology (see EEG asymmetry in depression above), it is
have generally found that patients likely to respond have interesting to explore whether this EEG variable might be
greater alpha power prior to treatment. This was first utilized in predicting antidepressant response. Initial observa-
shown in 20 in-patients with endogenous depression treated tions by Ulrich et al. were that the best baseline predictor
with amitriptyline or pirlindol, with responders having of response was alpha power over the left, but not right,
greater average alpha over the left occipital lobe (Ulrich occipital lobe (Ulrich et al., 1984). However, rather opposite
et al., 1984). This was further analysed suggesting that non- to this early finding, in a study of 53 patients treated with
response was associated with alpha power being less than fluoxetine (34 responders and 19 non-responders), non-
50% of total spectral power (Ulrich and Frick, 1986). This responders showed an alpha asymmetry indicative of overall
observation is supported by a similar, though non-significant, greater activation (i.e. less alpha power) of the right hemi-
effect in 29 patients treated with imipramine (Knott et al., sphere than the left, whereas responders did not (Bruder
1996), but there have also been negative findings with et al., 2001). This was, however, evident only in women
Ulrich et al. finding no baseline differences between respon- and not in men. Despite this, the group were able to repli-
ders and non-responders in 45 patients treated with either cate their finding in just 18 patients, only five of whom
clomipramine or maprotiline (Ulrich et al., 1988). Most were women, with no apparent effect of gender on the pre-
recently, a study investigating the predictive power of alpha diction of response, which had a sensitivity of 64% and
power in 18 patients treated with fluoxetine found this param- specificity of 71% (Bruder et al., 2008). The findings of
eter to have a sensitivity of 73% and a specificity of 58% this group appear to have been replicated by a Russian
(Bruder et al., 2008). team (based on the English translation of their abstract:
Three others studies have investigated the predictive Bochkarev et al., 2004).
potential of baseline alpha power using a particular analysis If the more recent data that a relative left frontal hypoac-
method referred to as ‘neurometric EEG mapping’. This map- tivity (i.e. higher alpha power) is associated with non-
ping is in two dimensions and involves EEG data being ini- response (Bochkarev et al., 2004; Bruder, 2008; Bruder
tially transformed into spectral power using the FFT (as in et al., 2001) are correct it is interesting to speculate what
others studies described above). The FFT data from whole this relates to. The pattern is the same as associated with
groups of patients are then z-transformed against population depression in general (see EEG asymmetry in depression
norms for age and log-transformed for normalization of dis- above) and hence it might just be an index of severity of
tribution. This method was first used to classify different psy- underlying trait psychological function (e.g. in relation to a
chiatric disorders in the 1980s (John et al., 1988). In their deficit of approach mechanism) and/or some biological dif-
naturalistic, prospective study of 54 patients with unipolar ference (e.g. in relation to 5-HT1A receptor function) that it is
or bipolar depression, Suffin and Emory recorded EEG difficult for antidepressants to overcome. It would be inter-
before starting six months of treatment with a tricyclic anti- esting to explore this EEG variable further in parallel with
depressant (TCA) or a SSRI. Of 35 responders, 30 (86%) had cognitive and other biological measures in patients undergo-
an excess of relative alpha power frontally and to a lesser ing treatment.
Alhaj et al. 1185

week of treatment into an ‘Antidepressant Treatment

EEG theta power
Response (ATR)’ index (Iosifescu et al., 2009). Using this
An alternative power spectrum band that has been explored methodology, in a study of 73 patients treated with escitalo-
extensively regarding its possible predictive value in relation pram, the ATR has been shown to have a 74% accuracy in
to antidepressant response is the theta band. Knott et al. predicting response (Leuchter et al., 2009).
reported that response to imipramine in 29 patients was
predicted by significantly less baseline theta than in non-
Loudness Dependency of Auditory Evoked Potentials
responders (Knott et al., 1996). This group have replicated
their findings in 70 male patients treated with paroxetine, 51 As discussed in Antidepressant effects on Loudness
of whom completed six weeks’ treatment. A negative corre- Dependency of Auditory Evoked Potentials (LDAEPs)
lation was found between baseline frontal theta and decrease above, LDAEPs have been used to study the serotonergic
in Hamilton Depression Rating Scale (HDRS) score. More function in humans (Hegerl and Juckel, 1993). It has also
recently, lower frontal theta absolute and relative power dif- been argued that the LDAEP has a significant promise as a
ferentiated responders from non-responders in 82 patients potential biomarker of antidepressant response (O’Neill
treated with venlafaxine or SSRIs with a sensitivity and et al., 2008). This is because LDAEPs’ ability to predict
specificity of 64% and 62% respectively (Iosifescu et al., antidepressant response has been replicated by more
2009). These findings are consistent with one neurometrics groups than for any other EEG method described above,
study in depressed patients where frontal theta excess and includes more patients than any other technique.
was identified in five of seven non-responders (Suffin and In addition there is a suggestion that it might be able to
Emory, 1995). differentially predict response to antidepressants from differ-
It has been hypothesized that activity/metabolic rate in ent pharmacological classes.
the rostral (pregenual) anterior cingulate cortex (rACC) pre- The first published study using the LDAEP to predict
dicts antidepressant response (Mayberg, 1997). This is based antidepressant response was not from the Juckle and
on observations that subjects with higher metabolic rates Hegerl group. Paige et al. studied the predictive power of
respond better to sleep deprivation (Ebert et al., 1994; the LDAEP (using just the P2 component) in 17 depressed
Smith et al., 1999; Wu et al., 1999) as well as paroxetine patients treated with a variety of antidepressants but mainly
(Brody et al., 1999; Saxena et al., 2003), sertraline SSRIs (12 of the 17). The response after 4–8 weeks’ treat-
(Buchsbaum et al., 1997) and venlafaxine (Davidson et al., ment correlated with the ASF-slope and all those who
2003). There is a close correlation between EEG rACC theta responded to an SSRI had a baseline ASF-slope greater
activity measured using LORETA and PET glucose metab- than the median (Paige et al., 1994). This group have sub-
olism (Pizzagalli et al., 2003). Three independent research sequently shown a similar result for the pharmacologically
groups have reported the use of the LORETA technique different antidepressant bupropion (Paige et al., 1995). The
to assess rACC theta activity in depressed patients pre- group of Juckel and Hegerl subsequently published a study
treatment. In the first of these Pizzagalli et al. (2001) of 29 patients treated for four weeks with SSRIs using the
investigated 18 depressed patients. All had been medication dipole source localization of the N1/P2 ERP complex tech-
free for at least four weeks prior to treatment with nortrip- nique (Gallinat et al., 2000). They divided the patients into
tyline with monitoring of plasma concentrations. Sixteen of two groups around the median value for the ASF-slope.
the 18 patients showed a response over 4–6 months, averag- Those with higher slopes showed a significantly greater
ing a 63% improvement in Beck Depressive Inventory (BDI) reduction in HDRS. Response rates (decrease in
score. All 18 patients were divided into those with a better HDRS  50%) for the high and low slope groups were 9
and those with a worse response around the median BDI out of 15 patients and 3 out of 14 respectively, and respon-
improvement. Comparisons between these two groups across ders overall had significantly greater ASF-slopes. The largest
all frequency bands revealed a significant difference only for replication of these findings comes from a Chinese group
theta and this was confined to rACC. The ‘better’ respon- who treated 100 patients with fluoxetine (Lee et al., 2005).
ders exhibited higher theta power than the ‘less-good’ Again patients were divided around the median ASF-slope
responders. Further, there was a significant correlation (N1/P2 peak to peak amplitude). After four weeks’ treat-
between rACC theta power and improvement in BDI ment those above the median had a 44% decrease in
(Pizzagalli et al., 2001). The findings of Pizzagalli et al. HDRS versus 34% for those below median.
have been replicated in a group of 20 patients randomized Intriguingly, the LDAEP may differentially predict
to either citalopram or reboxetine for four weeks (Mulert response to antidepressants with differing pharmacologies.
et al., 2007b). Again responders had higher rACC theta than Linka et al. have replicated the findings described above
non-responders with a Cohen’s effect size of 1.33, with a (though only in relation to N1 amplitude and not P2 or
positive correlation between rACC theta and decrease in N1/P2 peak to peak amplitudes) in 16 in-patients with
HDRS. Similar findings have also been reported for 72 sub- depression treated with the SSRI citalopram (Linka et al.,
jects from placebo controlled randomized trials of fluoxetine 2004). Conversely they report that there is an opposite
and venlafaxine with placebo responders not exhibiting relationship between ASF-slope and response to the
higher rACC activity (Korb et al., 2009). noradrenergic uptake blocker reboxetine (Linka et al.,
Recent work has examined the utility of combining base- 2005). This finding has been replicated by Juckel et al. in 35
line EEG theta parameters, prior to treatment with an anti- inpatients again treated with either citalopram or reboxetine
depressant, with early changes in EEG power during the first (Juckel et al., 2007).
1186 Journal of Psychopharmacology 25(9)

SSRIs) as the sensitivity of the predictive test. While the data

Combining various EEG/ERP parameters
as yet remain too speculative to suggest the use of the EEG
As described above, a number of different EEG techniques clinically irrespective of the treatment being considered, this is
have been explored as potential biomarkers of antidepressant clearly an area that requires further research. However, if the
response with several being of potential utility. Two groups currently extremely promising data in these areas are sup-
have attempted to combine more than one measure together ported by further replications and developments, this could
to increase the sensitivity and specificity in terms of response have a major impact possibly clinically but also particularly in
prediction. the development of novel pharmacological agents with the
Focusing on pre-treatment alpha spectral power, Bruder ability to pre-select cohorts of patients with likely high
et al. combined global alpha power (sensitivity 73% and response rates, increasing the power of early clinical trials.
specificity 58%) and alpha asymmetry (sensitivity 64% and
specificity 71%) giving an overall sensitivity of 83% and spe- Funding
cificity of 68% (Bruder et al., 2008). The relatively small
This research received no specific grant from any funding agency in
increase in sensitivity and specificity when combining these
the public, commercial, or not-for-profit sectors
measures may relate to the fact that a correlation (r ¼ 0.34)
existed between them. Of perhaps even more interest is a
report combining theta activity localized to rACC with References
LORETA and the LDAEP technique – two methods with Alhaj HA, Massey AE and McAllister-Williams RH (2007) A study
very different theoretical bases and that do not correlate of the neural correlates of episodic memory and HPA axis status
one with another (Mulert et al., 2007a). The combination of in drug-free depressed patients and healthy controls. J Psychiatr
the two techniques significantly discriminated responders Res 41(3–4): 295–304.
from non-responders in 20 patients randomized to either Alhaj HA, Massey AE and McAllister-Williams RH (2008) Effects of
cortisol on the laterality of the neural correlates of episodic
citalopram or reboxetine for four weeks with a large
memory. J Psychiatr Res 42(12): 971–981.
Cohen’s effect size of 1.41. Allen J, Urry H, Hitt S and Coan J (2004) The stability of resting
frontal electroencephalographic asymmetry in depression.
Conclusions Psychophysiology 41(2): 269–280.
Altshuler LL, Devinsky O, Post RM and Theodore W (1990)
EEG methods in general are often considered to not have the Depression, anxiety, and temporal lobe epilepsy. Laterality of
cache of other functional imaging techniques. However, they focus and symptoms. Arch Neurol 47(3): 284–288.
are inexpensive and easy to perform. This alone is an impor- Ancy J, Gangadhar BN and Janakiramaiah N (1996) ’Normal’ P300
tant starting point when considering whether the EEG could amplitude predicts rapid response to ECT in melancholia. J Affect
have utility in experimental medicine and the development of Disord 41(3): 211–215.
Anderer P, Saletu B and Pascual-Marqui RD (2000) Effect of the
new psychotropic medications. New and innovative methods
5-HT(1A) partial agonist buspirone on regional brain electrical
of utilizing the EEG are further enhancing its promise. EEG
activity in man: a functional neuroimaging study using low-reso-
and ERPs show potential as biomarkers for the illness and lution electromagnetic tomography (LORETA). Psychiatry Res
abnormalities found may have value as endophenotypes. In 100(2): 81–96.
depression, one of the most promising potential endopheno- Baillet S, Riera JJ, Marin G, Mangin JF, Aubert J and Garnero L
types is the relative laterality of EEG cortical activity, which (2001) Evaluation of inverse methods and head models for EEG
led to the generation of the ‘asymmetry hypothesis of depres- source localization using a human skull phantom. Phys Med Biol
sion’. The EEG has also proved of value in the exploration of 46(1): 77–96.
the pathophysiology underlying depression, for example Barbanoj MJ, Anderer P, Antonijoan RM, Torrent J, Saletu B and
examination of the effects of HPA axis abnormalities. Jane F (1994) Topographical pharmaco-EEG mapping of increas-
EEG abnormalities associated with depression have been ing doses of buspirone and its comparison with diazepam. Human
Psychopharm 9: 101–109.
utilized as a target for treatment. For example hemispheric
Bares M, Brunovsky M, Kopecek M, et al. (2007) Changes in QEEG
asymmetry is central to proposed mechanisms of rTMS. In prefrontal cordance as a predictor of response to antidepressants
terms of utilizing the EEG as a measure of the effects of psy- in patients with treatment resistant depressive disorder: a pilot
chotropics, it is clear that a number of early changes in a study. J Psychiatr Res 41(3–4): 319–325.
variety of parameters correlate strongly with antidepressant Bares M, Brunovsky M, Kopecek M, et al. (2008) Early reduction in
response (e.g. frontal theta cordance and early changes in prefrontal theta QEEG cordance value predicts response to ven-
theta activity especially localized to rostral ACC). Further, lafaxine treatment in patients with resistant depressive disorder.
the EEG is increasingly being suggested to be of potential Eur Psychiatry 23(5): 350–355.
benefit for the identification of patients likely to respond to Bares M, Brunovsky M, Novak T, et al. (2010) The change of pre-
medication. Baseline measures of EEG alpha power, alpha frontal QEEG theta cordance as a predictor of response to
bupropion treatment in patients who had failed to respond to
asymmetry, ACC theta power and LDAEP slope may all
previous antidepressant treatments. Eur Neuropsychopharmacol
have some utility. Recent interest is particularly focused on
20(7): 459–466.
combining multiple baseline EEG parameters, or baseline Bel N and Artigas F (1992) Fluvoxamine preferentially increases
parameters and early treatment effects, to increase accuracy. extracellular 5-hydroxytryptamine in the raphe nuclei: an
Whether or not a predictive EEG test would be useful clini- in vivo microdialysis study. Eur J Pharmacol 229(1): 101–103.
cally depends as much on the relative risk/benefit ratio of Bermpohl F, Fregni F, Boggio PS, et al. (2006) Effect of low-
particular treatments (with this being low for some such as frequency transcranial magnetic stimulation on an affective
Alhaj et al. 1187

go/no-go task in patients with major depression: role of stimula- (eds) The Asymmetrical Brain. Cambridge, MA: MIT Press,
tion site and depression severity. Psychiatry Res 141(1): 1–13. 565–615.
Bismark AW, Moreno FA, Stewart JL, et al. (2010) Polymorphisms Coan J and Allen J (2004) Frontal EEG asymmetry as a moderator
of the HTR1a allele are linked to frontal brain electrical asymme- and mediator of emotion. Biol Psychol 67(1–2): 7–50.
try. Biol Psychol 83(2): 153–158. Coburn KL, Lauterbach EC, Boutros NN, Black KJ, Arciniegas DB
Blackwood DH, Sharp CW, Walker MT, Doody GA, Glabus MF and Coffey CE (2006) The value of quantitative electroencepha-
and Muir WJ (1996) Implications of comorbidity for genetic stud- lography in clinical psychiatry: a report by the Committee on
ies of bipolar disorder: P300 and eye tracking as biological mar- Research of the American Neuropsychiatric Association.
kers for illness. Brit J Psychiatry 168(Suppl. 30): 85–92. J Neuropsychiatry Clin Neurosci 18(4): 460–500.
Blier P and de Montigny C (1994) Current advances in the treatment Compton RJ, Banich MT, Mohanty A, et al. (2003) Paying attention
of depression. Trends Pharmacol 15: 220–226. to emotion: an fMRI investigation of cognitive and emotional
Bochkarev VK, Avedisova AS and Liupaeva NV (2004) Placebo to stroop tasks. Cogn Affect Behav Neurosci 3(2): 81–96.
antidepressant effects ratio by electroencephalographic data. Cook IA, Leuchter AF, Morgan M, et al. (2002) Early changes in
Zhurnal Nevrologii i Psikhiatrii Imeni S S Korsakova 104(11): prefrontal activity characterize clinical responders to antidepres-
42–48. sants. Neuropsychopharmacology 27(1): 120–131.
Bogdanov NN and Bogdanov MB (1994) The role of 5-HT1A sero- Cook IA, Leuchter AF, Morgan ML, Stubbeman W, Siegman B and
tonin and D2 dopamine receptors in buspirone effects on cortical Abrams M (2005) Changes in prefrontal activity characterize clin-
electrical activity in rats. Neurosci Lett 177(1–2): 1–4. ical response in SSRI nonresponders: a pilot study. J Psychiatr
Brody AL, Saxena S, Silverman DH, et al. (1999) Brain metabolic Res 39(5): 461–466.
changes in major depressive disorder from pre- to post-treatment Davidson RJ (1998) Anterior electrophysiological asymmetries, emo-
with paroxetine. Psychiatry Res 91(3): 127–139. tion, and depression: conceptual and methodological conun-
Bruder GE, Tenke CE, Warner V, et al. (2005) drums. Psychophysiology 35(5): 607–614.
Electroencephalographic measures of regional hemispheric activ- Davidson RJ and Henriques J (2000) Regional brain function in sad-
ity in offspring at risk for depressive disorders. Biol Psychiatry ness and depression. In: Borod JC (ed) The Neuropsychology of
57(4): 328–335. Emotion. New York: Oxford University Press, 269–297.
Bruder GE, Sedoruk JP, Stewart JW, McGrath PJ, Quitkin FM and Davidson RJ, Irwin W, Anderle MJ and Kalin NH (2003) The neural
Tenke CE (2008) Electroencephalographic alpha measures predict substrates of affective processing in depressed patients treated
therapeutic response to a selective serotonin reuptake inhibitor
with venlafaxine. Am J Psychiatry 160(1): 64–75.
antidepressant: pre- and post-treatment findings. Biol Psychiatry
Debener S, Beauducel A, Nessler D, Brocke B, Heilemann H and
63(12): 1171–1177.
Kayser J (2000) Is resting anterior EEG alpha asymmetry a
Bruder GE, Stewart JW, Tenke CE, et al. (2001)
trait marker for depression? Findings for healthy adults and clin-
Electroencephalographic and perceptual asymmetry differences
ically depressed patients. Neuropsychobiology 41(1): 31–37.
between responders and nonresponders to an SSRI antidepres-
Debener S, Strobel A, Kurschner K, et al. (2002) Is auditory evoked
sant. Biol Psychiatry 49(5): 416–425.
potential augmenting/reducing affected by acute tryptophan
Bruder GE, Towey JP, Stewart JW, Friedman D, Tenke C and
depletion? Biol Psychol 59(2): 121–133.
Quitkin FM (1991) Event-related potentials in depression: influ-
de Visser SJ, van der Post JP, de Waal PP, Cornet F, Cohen AF and
ence of task, stimulus hemifield and clinical features on P3
van Gerven JM (2003) Biomarkers for the effects of
latency. Biol Psychiatry 30(3): 233–246.
benzodiazepines in healthy volunteers. Br J Clin Pharmacol
Bruder G, Fong R, Tenke CE, et al. (1997) Regional brain asymme-
55(1): 39–50.
tries in major depression with or without an anxiety disorder:
a quantitative electroencephalographic study. Biol Psychiatry Dierks T, Barta S, Demisch L, et al. (1999) Intensity dependence of
41(9): 939–948. auditory evoked potentials (AEPs) as biological marker for cere-
Buchsbaum MS, Wu J, Siegel BV, et al. (1997) Effect of sertraline on bral serotonin levels: effects of tryptophan depletion in healthy
regional metabolic rate in patients with affective disorder. Biol subjects. Psychopharmacology 146(1): 101–107.
Psychiatry 41(1): 15–22. Dimpfel W (2008) Pharmacological modulation of dopaminer-
Burt T, Lisanby SH and Sackeim HA (2002) Neuropsychiatric appli- gic brain activity and its reflection in spectral frequencies of
cations of transcranial magnetic stimulation: a meta analysis. Int J the rat electropharmacogram. Neuropsychobiology 58(3–4):
Neuropsychopharmacol 5(1): 73–103. 178–186.
Carozzo S, Fornaro S, Garbarino S, Saturno M and Sannita WG Dougherty DD and Rauch SL (2007) Brain correlates of antidepres-
(2006) From neuroscience to application in neuropharmacology: sant treatment outcome from neuroimaging studies in depression.
a generation of progress in electrophysiology. Clin EEG Neurosci Psychiatr Clin North Am 30(1): 91–103.
37(2): 121–134. Drevets WC and Raichle ME (1995) Positron emission tomographic
Carroll BJ, Curtis GC and Mendels J (1976) Cerebrospinal fluid and imaging studies of human emotional disorders. In: Gazzaniga MS
plasma free cortisol concentrations in depression. Psychol Med (ed) The Cognitive Neurosciences. Cambridge, MA: MIT Press,
6(2): 235–244. 1153–1164.
Chapotot F, Buguet A, Gronfier C and Brandenberger G (2001) Ebert D, Feistel H, Barocka A and Kaschka W (1994) Increased
Hypothalamo-pituitary-adrenal axis activity is related to the limbic blood flow and total sleep deprivation in major depression
level of central arousal: effect of sleep deprivation on the associ- with melancholia. Psychiatry Res 55(2): 101–109.
ation of high-frequency waking electroencephalogram with corti- Engels AS, Heller W, Mohanty A, et al. (2007) Specificity of regional
sol release. Neuroendocrin 73(5): 312–321. brain activity in anxiety types during emotion processing.
Clifford EM, Gartside SE, Umbers V, Cowen PJ, Hajos M and Psychophysiology 44(3): 352–363.
Sharp T (1998) Electrophysiological and neurochemical evidence Evans KC, Dougherty DD, Pollack MH and Rauch SL (2006) Using
that pindolol has agonist properties at the 5-HT(1A) autorecep- neuroimaging to predict treatment response in mood and anxiety
tor in vivo. Brit J Pharmacol 124(1): 206–212. disorders. Ann Clin Psychiatry 18(1): 33–42.
Coan JA and Allen JJB (2003) The state and trait nature of frontal Fawcett J and Barkin RL (1997) Efficacy issues with antidepressants.
EEG asymmetry in emotion. In: Hugdahl K and Davidson RJ J Clin Psychiatry 58(Suppl. 6): 32–39.
1188 Journal of Psychopharmacology 25(9)

Fell J, Klaver P, Elger CE and Fernandez G (2002) Suppression of Hesdorffer DC, Hauser WA, Annegers JF and Cascino G (2000)
EEG gamma activity may cause the attentional blink. Conscious Major depression is a risk factor for seizures in older adults.
Cogn 11(1): 114–122. Ann Neurol 47(2): 246–249.
Frodl T, Bokde AL, Scheuerecker J, et al. (2010) Functional connec- Hetzel G, Moeller O, Evers S, et al. (2005) The astroglial protein
tivity bias of the orbitofrontal cortex in drug-free patients with S100B and visually evoked event-related potentials before and
major depression. Biol Psychiatry 67(2): 161–167. after antidepressant treatment. Psychopharmacology 178(2–3):
Galderisi S, Mucci A, Bucci P, Mignone ML and Maj M (1996) 161–166.
Influence of moclobemide on cognitive functions of nine Hinrikus H, Suhhova A, Bachmann M, et al. (2009)
depressed patients: pilot trial with neurophysiological and neuro- Electroencephalographic spectral asymmetry index for detection
psychological indices. Neuropsychobiology 33(1): 48–54. of depression. Med Biol Eng Comput 47(12): 1291–1299.
Gallinat J, Bottlender R, Juckel G, et al. (2000) The loudness depen- Hjorth S and Auerbach SB (1994) Further evidence for the impor-
dency of the auditory evoked N1/P2-component as a predictor of tance of 5-HT1A autoreceptors in the action of selective serotonin
the acute SSRI response in depression. Psychopharmacology reuptake inhibitors. Eur J Pharmacol 260(2–3): 251–255.
148(4): 404–411. Holland RL, Wesnes K and Dietrich B (1994) Single dose human
Gershon AA, Dannon PN and Grunhaus L (2003) Transcranial mag- pharmacology of umespirone. Eur J Clin Pharmacol 46: 461–468.
netic stimulation in the treatment of depression. Am J Psychiatry Holthoff VA, Beuthien-Baumann B, Zundorf G, et al. (2004)
160(5): 835–845. Changes in brain metabolism associated with remission in unipo-
Gibbons JL and McHugh PR (1962) Plasma cortisol in depressive lar major depression. Acta Psychiatr Scand 110(3): 184–194.
illness. J Psychiatr Res 1: 162–171. Holtzheimer III PE, Russo J and Avery DH (2001) A meta-analysis
Grin-Yatsenko VA, Baas I, Ponomarev VA and Kropotov JD (2010) of repetitive transcranial magnetic stimulation in the treatment of
Independent component approach to the analysis of EEG record- depression. Psychopharmacol Bull 35(4): 149–169.
ings at early stages of depressive disorders. Clin Neurophysiol Hsu FC, Garside MJ, Massey AE and McAllister-Williams RH
121(3): 281–289. (2003) Effects of a single dose of cortisol on the neural correlates
Gronfier C, Chapotot F, Weibel L, Jouny C, Piquard F and of episodic memory and error processing in healthy volunteers.
Brandenberger G (1998) Pulsatile cortisol secretion and EEG Psychopharmacology 167(4): 431–442.
delta waves are controlled by two independent but synchronized Iosifescu DV, Greenwald S, Devlin P, et al. (2009) Frontal EEG
generators. Am J Physiol 275(1 Pt 1): E94–100. predictors of treatment outcome in major depressive disorder.
Gruzelier J (2009) A theory of alpha/theta neurofeedback, creative Eur Neuropsychopharmacol 19(11): 772–777.
performance enhancement, long distance functional connectivity Ising M, Horstmann S, Kloiber S, et al. (2007) Combined dexameth-
and psychological integration. Cogn Process 10(Suppl. 1): asone/corticotropin releasing hormone test predicts treatment
S101–S109. response in major depression – a potential biomarker? Biol
Guidi M, Scarpino O and Angeleri F (1989) Topographic EEG and Psychiatry 62(1): 47–54.
flash visual evoked potentials in elderly subjects, depressed John ER and Prichep LS (2006) The relevance of QEEG to the eval-
patients, and demented patients. Psychiatry Res 29(3): 403–406. uation of behavioral disorders and pharmacological interventions.
Hagemann D (2004) Individual differences in anterior EEG asymme- Clin EEG Neurosci 37(2): 135–143.
try: methodological problems and solutions. Biol Psychol 67(1–2): John ER, Prichep LS, Fridman J and Easton P (1988) Neurometrics:
157–182. computer-assisted differential diagnosis of brain dysfunctions.
Handy TC (2004) Event-Related Potentials: A Methods Handbook. Science 239(4836): 162–169.
Cambridge, MA: MIT Press. Jones NA, Field T, Fox NA, Lundy B and Davalos M (1997) EEG
Hansen ES, Prichep LS, Bolwig TG and John ER (2003) Quantitative activation in 1-month-old infants of depressed mothers. Dev
electroencephalography in OCD patients treated with paroxetine. Psychopathol 9(3): 491–505.
Clin Electroencephalogr 34(2): 70–74. Juckel G, Hegerl U, Molnar M, Csepe V and Karmos G (1999)
Hansenne M and Ansseau M (1999) P300 event-related potential and Auditory evoked potentials reflect serotonergic neuronal activity
serotonin-1A activity in depression. Eur Psychiatry 14(3): – a study in behaving cats administered drugs acting on 5-HT1A
143–147. autoreceptors in the dorsal raphe nucleus.
Harden CL (2002) The co-morbidity of depression and epilepsy: epi- Neuropsychopharmacology 21(6): 710–716.
demiology, etiology, and treatment. Neurology 59(6 Suppl. 4): Juckel G, Pogarell O, Augustin H, et al. (2007) Differential prediction
S48–S55. of first clinical response to serotonergic and noradrenergic anti-
Hegerl U and Juckel G (1993) Intensity dependence of auditory depressants using the loudness dependence of auditory evoked
evoked potentials as an indicator of central serotonergic neuro- potentials in patients with major depressive disorder. J Clin
transmission: a new hypothesis. Biol Psychiatry 33(3): 173–187. Psychiatry 68(8): 1206–1212.
Hegerl U, Gallinat J and Juckel G (2001) Event-related potentials. Kawasaki T, Tanaka S, Wang J, Hokama H and Hiramatsu K (2004)
Do they reflect central serotonergic neurotransmission and do Abnormalities of P300 cortical current density in unmedicated
they predict clinical response to serotonin agonists? J Affect depressed patients revealed by LORETA analysis of event-related
Disord 62(1–2): 93–100. potentials. Psychiatry Clin Neurosci 58(1): 68–75.
Henriques JB and Davidson RJ (1990) Regional brain electrical Kentgen L, Tenke C, Pine D, Fong R, Klein R and Bruder G (2000)
asymmetries discriminate between previously depressed and Electroencephalographic asymmetries in adolescents with major
healthy control subjects. J Abnorm Psychol 99(1): 22–31. depression: influence of comorbidity with anxiety disorders.
Henriques JB and Davidson R (1991) Left frontal hypoactivation in J Abnorm Psychol 109(4): 797–802.
depression. J Abnorm Psychol 100(4): 535–545. Killgore WD, Gruber SA and Yurgelun-Todd DA (2007) Depressed
Herrington JD, Heller W, Mohanty A, et al. (2010) Localization of mood and lateralized prefrontal activity during a Stroop task in
asymmetric brain function in emotion and depression. adolescent children. Neurosci Lett 416(1): 43–48.
Psychophysiology 47(3): 442–454. Kimbrell TA, Little JT, Dunn RT, et al. (1999) Frequency depen-
Herrmann CS, Frund I and Lenz D (2010) Human gamma-band dence of antidepressant response to left prefrontal repetitive tran-
activity: a review on cognitive and behavioral correlates and net- scranial magnetic stimulation (rTMS) as a function of baseline
work models. Neurosci Biobehav Rev 34(7): 981–992. cerebral glucose metabolism. Biol Psychiatry 46(12): 1603–1613.
Alhaj et al. 1189

Kirsch I, Deacon BJ, Huedo-Medina TB, Scoboria A, Moore TJ and Linka T, Sartory G, Bender S, Gastpar M and Muller BW (2007) The
Johnson BT (2008) Initial severity and antidepressant benefits: a intensity dependence of auditory ERP components in unmedi-
meta-analysis of data submitted to the Food and Drug cated patients with major depression and healthy controls.
Administration. PLoS Med 5(2): e45. An analysis of group differences. J Affect Disord 103(1–3):
Klimesch W (1999) EEG alpha and theta oscillations reflect cognitive 139–145.
and memory performance: a review and analysis. Brain Res Brain Marrosu F, Fornal CA, Metzler CW and Jacobs BL (1996) 5-HT1A
Res Rev 29(2–3): 169–195. agonists induce hippocampal theta activity in freely moving cats:
Knott VJ (2000) Quantitative EEG methods and measures in human role of presynaptic 5-HT1A receptors. Brain Res 739(1–2):
psychopharmacological research. Hum Psychopharmacol 15(7): 192–200.
479–498. Massey AE, Marsh VR and McAllister-Williams RH (2004) Lack of
Knott VJ, Telner JI, Lapierre YD, Browne M and Horn ER (1996) effect of tryptophan depletion on the loudness dependency of
Quantitative EEG in the prediction of antidepressant response to auditory event related potentials in healthy volunteers. Biol
imipramine. J Affect Disord 39(3): 175–184. Psychol 65(2): 137–145.
Knyazev GG (2007) Motivation, emotion, and their inhibitory con- Mayberg HS (1997) Limbic-cortical dysregulation: a proposed model
trol mirrored in brain oscillations. Neurosci Biobehav Rev 31(3): of depression. f Neuropsychiatry Clin Neurosci 9(3): 471–481.
377–395. Mayberg HS, Liotti M, Brannan SK, et al. (1999) Reciprocal limbic-
Koles ZJ (1998) Trends in EEG source localization. cortical function and negative mood: converging PET findings in
Electroencephalogr Clin Neurophysiol 106(2): 127–137. depression and normal sadness. Am J Psychiatry 156(5): 675–682.
Korb AS, Cook IA, Hunter AM and Leuchter AF (2008) Brain elec- McAllister-Williams RH (2008) Misinterpretation of randomized
trical source differences between depressed subjects and healthy trial evidence: do antidepressants work? Br J Hosp Med (Lond)
controls. Brain Topogr 21(2): 138–146. 69(5): 246–247.
Korb AS, Hunter AM, Cook IA and Leuchter AF (2009) Rostral McAllister-Williams RH and Massey AE (2003) EEG effects of bus-
anterior cingulate cortex theta current density and response to pirone and pindolol: a method of examining 5-HT(1A) receptor
antidepressants and placebo in major depression. Clin function in humans. Psychopharmacology 166(3): 284–293.
Neurophysiol 120(7): 1313–1319. McAllister-Williams RH and Rugg MD (2002) Effects of repeated
Korostenskaja M and Kahkonen S (2009) What do ERPs and ERFs cortisol administration on brain potential correlates of episodic
reveal about the effect of antipsychotic treatment on cognition in memory retrieval. Psychopharmacology 160(1): 74–83.
McAllister-Williams RH, Marsh VR and Massey AE (2004)
schizophrenia? Curr Pharm Des 15(22): 2573–2593.
Increased sensitivity of somatodendritic 5-HT1A autorecptors in
Kubicki S and Herrmann WM (1996) The future of computer-
depression. J Psychopharm 18(3 Suppl): A17.
assisted investigation of the polysomnogram: sleep microstruc-
McAllister-Williams RH, Massey AE and Fairchild G (2007)
ture. J Clin Neurophysiol 13(4): 285–294.
Repeated cortisol administration attenuates the EEG response
Kuo CC and Tsai JF (2010) Cordance or Antidepressant Treatment
to buspirone in healthy volunteers: evidence for desensitization
Response (ATR) index? Psychiatry Res 180(1): 60.
of the 5-HT1A autoreceptor. J Psychopharmacol 21(8): 826–832.
Kwon JS, Youn T and Jung HY (1996) Right hemisphere abnormal-
McAllister-Williams RH, Hanson A, Alhaj HA, Marsh VR, Massey
ities in major depression: quantitative electroencephalographic
AE and Pariante CM (2009) Citalopram modulation of the effects
findings before and after treatment. J Affect Disord 40(3):
of cortisol on cortical activity in healthy male subjects. Biol
169–173.
Psychiatry 65(8S): 238S.
Laufs H, Kleinschmidt A, Beyerle A, et al. (2003) EEG-correlated
Michel CM, Murray MM, Lantz G, Gonzalez S, Spinelli L and
fMRI of human alpha activity. Neuroimage 19(4): 1463–1476.
Grave DP (2004) EEG source imaging. Clin Neurophysiol
Lee TW, Yu YW, Chen TJ and Tsai SJ (2005) Loudness dependence
115(10): 2195–2222.
of the auditory evoked potential and response to antidepressants Miller JM, Oquendo MA, Ogden RT, Mann JJ and Parsey RV (2008)
in Chinese patients with major depression. J Psychiatry Neurosci Serotonin transporter binding as a possible predictor of one-year
30(3): 202–205. remission in major depressive disorder. J Psychiatr Res 42(14):
Leuchter AF, Cook IA, Lufkin RB, et al. (1994) Cordance: a new 1137–1144.
method for assessment of cerebral perfusion and metabolism Monk CS and Nelson CA (2002) The effects of hydrocortisone on
using quantitative electroencephalography. Neuroimage 1(3): cognitive and neural function: a behavioral and event-related
208–219. potential investigation. Neuropsychopharmacology 26(4): 505–519.
Leuchter AF, Uijtdehaage SH, Cook IA, O’Hara R and Mandelkern Moore P, Landolt HP, Seifritz E, et al. (2000) Clinical and physio-
M (1999) Relationship between brain electrical activity and cor- logical consequences of rapid tryptophan depletion.
tical perfusion in normal subjects. Psychiatry Res 90(2): 125–140. Neuropsychopharmacology 23(6): 601–622.
Leuchter AF, Cook IA, Witte EA, Morgan M and Abrams M (2002) Morgan ML, Witte EA, Cook IA, Leuchter AF, Abrams M and
Changes in brain function of depressed subjects during treatment Siegman B (2005) Influence of age, gender, health status, and
with placebo. Am J Psychiatry 159(1): 122–129. depression on quantitative EEG. Neuropsychobiology 52(2): 71–76.
Leuchter AF, Cook IA, Marangell LB, et al. (2009) Comparative Mottaghy FM, Keller CE, Gangitano M, et al. (2002) Correlation of
effectiveness of biomarkers and clinical indicators for predicting cerebral blood flow and treatment effects of repetitive transcranial
outcomes of SSRI treatment in Major Depressive Disorder: magnetic stimulation in depressed patients. Psychiatry Res 115(1–2):
results of the BRITE-MD study. Psychiatry Res 169(2): 124–131. 1–14.
Linka T, Muller BW, Bender S and Sartory G (2004) The intensity Mucci A, Volpe U, Merlotti E, Bucci P and Galderisi S (2006)
dependence of the auditory evoked N1 component as a predictor Pharmaco-EEG in psychiatry. Clin EEG Neurosci 37(2): 81–98.
of response to Citalopram treatment in patients with major Mulert C, Juckel G, Brunnmeier M, et al. (2007a) Prediction of treat-
depression. Neurosci Lett 367(3): 375–378. ment response in major depression: integration of concepts.
Linka T, Muller BW, Bender S, Sartory G and Gastpar M (2005) The J Affect Disord 98(3): 215–225.
intensity dependence of auditory evoked ERP components pre- Mulert C, Juckel G, Brunnmeier M, et al. (2007b) Rostral anterior
dicts responsiveness to reboxetine treatment in major depression. cingulate cortex activity in the theta band predicts response to
Pharmacopsychiatry 38(3): 139–143. antidepressive medication. Clin EEG Neurosci 38(2): 78–81.
1190 Journal of Psychopharmacology 25(9)

Mulert C, Pogarell O, Juckel G, et al. (2004) The neural basis of the cingulate cortex: an EEG/PET study of normal and depressed
P300 potential. Focus on the time-course of the underlying corti- subjects. Psychophysiology 40(6): 939–949.
cal generators. Eur Arch Psychiatry Clinl Neurosci 254(3): Pollock VE and Schneider LS (1990) Quantitative, waking EEG
190–198. research on depression. Biol Psychiatry 27(7): 757–780.
Murasaki M, Miura S, Ishigooka J, Ishii Y, Takahashi A and Prichep LS, Mas F, Hollander E, et al. (1993) Quantitative electro-
Fukuyama Y (1989) Phase I study of a new antianxiety drug, encephalographic subtyping of obsessive-compulsive disorder.
buspirone. Prog Neuropsychopharmacol Biol Psychiatry 13(1–2): Psychiatry Res 50(1): 25–32.
137–144. Pruus K, Skrebuhhova-Malmros T, Rudissaar R, Matto V and
Murphy FC, Nimmo-Smith I and Lawrence AD (2003) Functional Allikmets L (2000) 5-HT1A receptor agonists buspirone and
neuroanatomy of emotions: a meta-analysis. Cogn Affect Behav gepirone attenuate apomorphine-induced aggressive behaviour
Neurosci 3(3): 207–233. in adult male Wistar rats. J Physiol Pharmacol 51(4 Pt 2):
Niedermeyer E and Da Silva FHL (2004) Electroencephalography: 833–846.
Basic Principles, Clinical Applications and Related Fields, 5th ed. Putnam KM and McSweeney LB (2008) Depressive symptoms and
Philadelphia: Lippincott Williams and Wilkins. baseline prefrontal EEG alpha activity: a study utilizing
Nitz DA and McNaughton BL (1999) Hippocampal EEG and unit Ecological Momentary Assessment. Biol Psychol 77(2): 237–240.
activity responses to modulation of serotonergic median raphe Reid SA, Duke LM and Allen JJ (1998) Resting frontal electroen-
neurons in the freely behaving rat. Learn Mem 6(2): 153–167. cephalographic asymmetry in depression: inconsistencies suggest
Nunez PL and Srinivasan R (2006) Electric Fields of the Brain: The the need to identify mediating factors. Psychophysiology 35(4):
Neurophysics of EEG, 2nd ed. New York: Oxford University 389–404.
Press. Reilly JG, McTavish SF and Young AH (1997) Rapid depletion of
O’Donnell BF, Vohs JL, Hetrick WP, Carroll CA and Shekhar A plasma tryptophan: a review of studies and experimental method-
(2004) Auditory event-related potential abnormalities in bipolar ology. J Psychopharm 11(4): 381–392.
disorder and schizophrenia. Int J Psychophysiol 53(1): 45–55. Rijnders HJ and Slangen JL (1993) The discriminative stimulus prop-
Ollikainen JO, Vauhkonen M, Karjalainen PA and Kaipio JP (2000) erties of buspirone involve dopamine-2 receptor antagonist activ-
Effects of electrode properties on EEG measurements and a ity. Psychopharmacology (Berl) 111(1): 55–61.
related inverse problem. Med Eng Phys 22(8): 535–545. Saito A, Kinoshita T, Okajima Y, Yagyu T, Saito M and
O’Neill BV, Croft RJ and Nathan PJ (2008) The loudness dependence Schollnhammer G (1993) Quantitative pharmaco-EEG study of
of the auditory evoked potential (LDAEP) as an in vivo bio-
ipsapirone (BAY q 7821) in healthy volunteers. Jap J Neuropharm
marker of central serotonergic function in humans: rationale,
15: 359–373.
evaluation and review of findings. Hum Psychopharmacol 23(5):
Saletu B, Grunberger J and Linzmayer L (1986) On central effects of
355–370.
serotonin re-uptake inhibitors: quantitative EEG and psychomet-
Padberg F, Zwanzger P, Thoma H, et al. (1999) Repetitive transcra-
ric studies with sertraline and zimelidine. J Neural Trans 67(3–4):
nial magnetic stimulation (rTMS) in pharmacotherapy-refractory
241–266.
major depression: comparative study of fast, slow and sham
Saletu B, Anderer P, Saletu-Zyhlarz GM, Arnold O and Pascual-
rTMS. Psychiatry Res 88(3): 163–171.
Marqui RD (2002) Classification and evaluation of the pharma-
Paige SR, Fitzpatrick DF, Kline JP, Balogh SE and Hendricks SE
codynamics of psychotropic drugs by single-lead pharmaco-EEG,
(1994) Event-related potential amplitude/intensity slopes predict
EEG mapping and tomography (LORETA). Methods Find Exp
response to antidepressants. Neuropsychobiology 30(4): 197–201.
Clin Pharmacol 24(Suppl. C): 97–120.
Paige SR, Hendricks SE, Fitzpatrick DF, Balogh S and Burke WJ
Sauseng P, Klimesch W, Gruber W, Doppelmayr M, Stadler W and
(1995) Amplitude/intensity functions of auditory event-related
potentials predict responsiveness to bupropion in major depres- Schabus M (2002) The interplay between theta and alpha oscilla-
sive disorder. Psychopharm Bull 31(2): 243–248. tions in the human electroencephalogram reflects the transfer of
Pascual-Leone A, Catala MD and Pascual-Leone PA (1996) information between memory systems. Neurosci Lett 324(2):
Lateralized effect of rapid-rate transcranial magnetic stimulation 121–124.
of the prefrontal cortex on mood. Neurology 46(2): 499–502. Savitz J, Lucki I and Drevets WC (2009) 5-HT(1A) receptor function
Pascual-Marqui RD, Michel CM and Lehmann D (1994) Low reso- in major depressive disorder. Prog Neurobiol 88(1): 17–31.
lution electromagnetic tomography: a new method for localizing Saxena S, Brody AL, Ho ML, Zohrabi N, Maidment KM and
electrical activity in the brain. Int J Psychophysiol 18(1): 49–65. Baxter Jr LR (2003) Differential brain metabolic predictors of
Perlis RH (2007) Pharmacogenetic studies of antidepressant response: response to paroxetine in obsessive-compulsive disorder versus
how far from the clinic? Psychiatr Clin North Am 30(1): 125–138. major depression. Am J Psychiatry 160(3): 522–532.
Phillips MA, Oxtoby EK, Langley RW, Bradshaw CM and Szabadi Scherg M and Von Cramon D (1986) Evoked dipole source potentials
E (2000) Effects of acute tryptophan depletion on prepulse inhi- of the human auditory cortex. Electroencephalogr Clin
bition of the acoustic startle (eyeblink) response and the N1/P2 Neurophysiol 65(5): 344–360.
auditory evoked response in man. J Psychopharm 14(3): 258–265. Schmidt LA, Fox NA, Goldberg MC, Smith CC and Schulkin J
Phillips C, Rugg MD and Fristont KJ (2002) Systematic regulariza- (1999) Effects of acute prednisone administration on memory,
tion of linear inverse solutions of the EEG source localization attention and emotion in healthy human adults.
problem. Neuroimage 17(1): 287–301. Psychoneuroendocrinology 24(4): 461–483.
Picton TW, Bentin S, Berg P, et al. (2000) Guidelines for using human Serretti A, Artioli P and Quartesan R (2005) Pharmacogenetics in the
event-related potentials to study cognition: recording standards treatment of depression: pharmacodynamic studies.
and publication criteria. Psychophysiology 37(2): 127–152. Pharmacogenet Genomics 15(2): 61–67.
Pizzagalli D, Pascual-Marqui RD, Nitschke JB, et al. (2001) Anterior Siegle GJ, Condray R, Thase ME, Keshavan M and Steinhauer SR
cingulate activity as a predictor of degree of treatment response in (2010) Sustained gamma-band EEG following negative words in
major depression: evidence from brain electrical tomography depression and schizophrenia. Int J Psychophysiol 75(2): 107–118.
analysis. Am J Psychiatry 158(3): 405–415. Siepmann M, Krause S, Joraschky P, Muck-Weymann M and Kirch
Pizzagalli DA, Oakes TR and Davidson RJ (2003) Coupling of theta W (2002) The effects of St John’s wort extract on heart rate var-
activity and glucose metabolism in the human rostral anterior iability, cognitive function and quantitative EEG: a comparison
Alhaj et al. 1191

with amitriptyline and placebo in healthy men. Br J Clin Ulrich G, Haug HJ, Stieglitz RD and Fahndrich E (1988) Are there
Pharmacol 54(3): 277–282. distinct biochemical subtypes of depression? EEG characteristics
Smit DJ, Posthuma D, Boomsma DI and Geus EJ (2005) Heritability of clinically defined on-drug responders and non-responders.
of background EEG across the power spectrum. Psychophysiology J Affect Disord 15(2): 181–185.
42(6): 691–697. Valle M, Barbanoj MJ and Jane F (2002) Pharmaco-electroencepha-
Smith GS, Reynolds III CF, Pollock B, et al. (1999) Cerebral glucose lography and pharmacokinetic/pharmacodynamic modeling in
metabolic response to combined total sleep deprivation and anti- basic research: focus on human pharmacology. Methods Find
depressant treatment in geriatric depression. Am J Psychiatry Exp Clin Pharmacol 24(Suppl. D): 129–137.
156(5): 683–689. Van der Stelt O and Belger A (2007) Application of electroencepha-
Souza VB, Muir WJ, Walker MT, et al. (1995) Auditory P300 event- lography to the study of cognitive and brain functions in schizo-
related potentials and neuropsychological performance in schizo- phrenia. Schizophr Bull 33(4): 955–970.
phrenia and bipolar affective disorder. Biol Psychiatry 37(5): Vertes RP (1982) Brain stem generation of the hippocampal EEG.
300–310. Prog Neurobiol 19(3): 159–186.
Sternbach HA, Extein I, Sweeney DR, Gold MS and Pottash AL Vertes RP, Kinney GG, Kocsis B and Fortin WJ (1994)
(1983) Cortisol secretion and urinary MHPG in unipolar depres- Pharmacological suppression of the median raphe nucleus with
sion. Int J Psychiatry Med 13(4): 261–266. serotonin1A agonists, 8-OH-DPAT and buspirone, produces hip-
Stubbeman WF, Leuchter AF, Cook IA, et al. (2004) Pretreatment pocampal theta rhythm in the rat. Neuroscience 60(2): 441–451.
neurophysiologic function and ECT response in depression. Volpe U, Mucci A, Bucci P, Merlotti E, Galderisi S and Maj M
J ECT 20(3): 142–144. (2007) The cortical generators of P3a and P3b: a LORETA
Suffin SC and Emory WH (1995) Neurometric subgroups in atten- study. Brain Res Bull 73(4–6): 220–230.
tional and affective disorders and their association with pharma- Wager TD, Phan KL, Liberzon I and Taylor SF (2003) Valence,
cotherapeutic outcome. Clin Electroencephalogr 26(2): 76–83. gender, and lateralization of functional brain anatomy in emo-
Thibodeau R, Jorgensen RS and Kim S (2006) Depression, anxiety, tion: a meta-analysis of findings from neuroimaging. Neuroimage
and resting frontal EEG asymmetry: a meta-analytic review. 19(3): 513–531.
J Abnorm Psychol 115(4): 715–729. Wisniewski GS, McAllister-Williams RH, Jakitowicz J, Nowicki Z,
Tops M, Wijers AA, van Staveren AS, et al. (2005) Acute cortisol Majkowicz M and Trzebiatowska IA (2008) Effects of antidepres-
administration modulates EEG alpha asymmetry in volunteers: sant drugs on electroencephalography alpha power: importance
relevance to depression. Biol Psychol 69(2): 181–193. of study duration. Biol Psychiatry 64(10): e5–e6.
Towers DN and Allen JJ (2009) A better estimate of the internal Wrobel A (2000) Beta activity: a carrier for visual attention. Acta
consistency reliability of frontal EEG asymmetry scores. Neurobiol Exp (Wars) 60(2): 247–260.
Psychophysiology 46(1): 132–142. Wu J, Buchsbaum MS, Gillin JC, et al. (1999) Prediction of antide-
Ulrich G and Frick K (1986) A new quantitative approach to the pressant effects of sleep deprivation by metabolic rates in the
assessment of stages of vigilance as defined by spatiotemporal ventral anterior cingulate and medial prefrontal cortex. Am J
EEG patterning. Percept Mot Skills 62(2): 567–576. Psychiatry 156(8): 1149–1158.
Ulrich G, Renfordt E, Zeller G and Frick K (1984) Yao D (2001) A method to standardize a reference of scalp EEG
Interrelation between changes in the EEG and psychopathology recordings to a point at infinity. Physiol Measur 22(4): 693–711.
under pharmacotherapy for endogenous depression. A contribu- Zyss T (2009) May depression be a form of epilepsy? Some remarks
tion to the predictor question. Pharmacopsychiatry 17(6): on the bioelectric nature of depression. Med Hypotheses 73(5):
178–183. 746–752.