Preliminary communication

Resting electroencephalographic correlates of the clinical responseto repetitive transcranial magnetic stimulation: A preliminarycomparison between unipolar and bipolar depression

Agata Woźniak-Kwaśniewska 1,a, David Szekely 2,a,b, Sylvain Harquel c,d, Thierry Bougerol b,Olivier David a,n

a EA 4617 Fonctions Cérébrales et Neuromodulation, Grenoble Institut des Neurosciences, Université Joseph Fourier, Grenoble, Franceb Clinique Universitaire de Psychiatrie, Pôle Psychiatrie Neurologie, Centre Hospitalier Universitaire, Grenoble, Francec UMR 5105 UPMF CNRS, Laboratoire Psychologie Neurocognition, Saint Martin d'Hères, Franced UMS IRMaGe, Grenoble, France

a r t i c l e i n f o

Article history:Received 10 November 2014Received in revised form15 April 2015Accepted 15 April 2015Available online 24 April 2015

Keywords:rTMSEEGMood disordersDorso-lateral prefrontal cortexSource localisationStatistical parametric mapping

a b s t r a c t

Background: Major depressive disorder (MDD) and bipolar disorder (BP) are two different types of mooddisorders, sometimes difficult to distinguish from their depressive symptoms, and for which repetitivetranscranial magnetic stimulation (rTMS) of the dorsolateral prefrontal cortex (DLPFC) has beenproposed to treat refractory patients. Here we studied whether the electroencephalogram (EEG) at restcould be used to predict the therapeutic response to left DLPFC 10 Hz rTMS, and to which extent BP andMDD patients show similar correlation between the clinical response and the cortical networks at rest.Methods: Eight MDD (6 females) and 10 BP patients (6 females) were included. The rTMS therapyconsisted of 10 to 20 neuronavigated sessions, with 2000 pulses continuously applied at 120% motorthreshold for each session. RTMS sessions at the beginning, middle and end of the therapy wereperformed while recording EEG signals. EEG spectral power was partitioned using the commonphysiological frequency bands and was statistically analysed at the scalp level and after cortical sourcereconstruction.Results: We found significantly higher power in theta and beta bands in BP patients than in MDDpatients, mainly localised in the prefrontal cortex. In addition, responders showed higher power in deltaand theta bands in parietal regions and weaker frontal alpha power, when compared to non-responders.Discussion: These preliminary findings on a small cohort suggest that pre-treatment EEG oscillatorypatterns may have some predictive value regarding rTMS therapy, both for MDD and BP disorders.

& 2015 Elsevier B.V. All rights reserved.

1. Introduction

Repetitive transcranial magnetic stimulation (rTMS) of dorsolateralprefrontal cortex (DLPFC) has been shown to be an effective

therapeutic treatment for major depressive disorder (MDD)(Carpenter et al., 2012; Fitzgerald et al., 2006; George et al., 2000,1995; Pascual-Leone et al., 1996). Studies on rTMS in bipolar depres-sion (BP) are less numerous (Dell’Osso et al., 2009; Dolberg et al.,2002; Tamas et al., 2007). Neurobiological mechanisms underlying thetherapeutic response to rTMS are supposed to involve brain areasremote from the targeted site, thereby altering oscillatory activity,plasticity and connectivity of neural networks at a large scale. Inparticular, it is likely that the limbic-cortical network, composed ofcortical (dorsolateral, medial and ventral prefrontal cortices) andlimbic (hippocampus and amygdala) structures (Li et al., 2010), ismodulated by rTMS of DLPFC.

EEG brought important neurophysiological information aboutdepression and various studies have found that typically 20–40% ofdepressed patients have EEG abnormalities, with several character-istics and controversial patterns (Coburn et al., 2006). Moreover, upto 80% of psychiatric patients present various EEG abnormalities on

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/jad

Journal of Affective Disorders

http://dx.doi.org/10.1016/j.jad.2015.04.0290165-0327/& 2015 Elsevier B.V. All rights reserved.

Abbreviations: ANOVA, analysis of variance; BP, bipolar disorder; CGI, ClinicalGlobal Impression; BDI, Beck Depression Inventory; DLPFC, dorso-lateral prefrontalcortex; DSM-IV, Diagnostic and Statistical Manual of Mental Disorder 4th edition;EEG, electroencephalography; FFT, fast Fourier transform; ICA, independent com-ponent analysis; MADRS, Montgomery Asberg Depression Rate Scale; MDD, majordepressive disorder; MRI, magnetic resonance imaging; RMT, resting motorthreshold; rTMS, repetitive transcranial magnetic stimulation; TMS, transcranialmagnetic stimulation; YMRS, Young Mania Rating Scale

n Corresponding author. Tel.: þ33 4 56 52 05 86; fax: þ33 4 56 52 05 98.E-mail address: Olivier.David@inserm.fr (O. David).1 Present address: Universität zu Lübeck, Institute of Physiology, Ratzeburger

Allee 160, D-23562 Luebeck.2 Present address: Clinique de Psychiatrie, Centre Hospitalier Princesse Grace,

1 avenue Pasteur, BP 489 MC 98012 Cedex, Principauté de Monaco.

Journal of Affective Disorders 183 (2015) 15–21

contrary to only 10% of healthy subjects (Begić et al., 2011; Coburnet al., 2006). Increased frontal EEG asymmetry, as measured byrelatively higher left than right alpha band activity, has beenproposed to characterise depressed individuals (Allen et al., 2004;Debener et al., 2000; Vuga et al., 2006) and to be a state-marker ofdepression (Carpenter et al., 2012; Debener et al., 2000; Fitzgeraldet al., 2006; George et al., 2000, 1995; Pascual-Leone et al., 1996). Inaddition, many studies have shown the association of theta altera-tions with treatment response to antidepressants, although thesefindings are inconsistent across studies (Dell’Osso et al., 2009;Dolberg et al., 2002; Heikman et al., 2001; Iosifescu et al., 2009;Knott et al., 2001, 1996; Tamas et al., 2007). In general, themodulations in frontal theta band have been interpreted as reflect-ing altered activity in the anterior cingulate cortex implicated inemotional regulations (Iosifescu, 2011; Li et al., 2010).

The main interest of EEG recordings in a clinical setting fordepression is to try to predict the response to therapy. In thiscontext, inter-hemispheric asymmetry of occipital alpha and pre-frontal theta bands have been reported as a valuable pre-treatment EEG features to differentiate responders from non-responders (Bruder et al., 2008; Carpenter et al., 2012; Fitzgeraldet al., 2006; George et al., 2000, 1995; Iosifescu et al., 2009;Pascual-Leone et al., 1996). Alpha asymmetry at baseline was alsoidentified as a possible discriminator: responders showed greateralpha power over right than over left hemisphere, whereas non-responders tended to show the opposite pattern (Bruder et al.,2008, 2001; Dell’Osso et al., 2009; Dolberg et al., 2002; Knott et al.,1996; Tamas et al., 2007; Tenke et al., 2011; Ulrich et al., 1984).However, only one study was published so far on the predictivevalue of EEG to 10 Hz rTMS therapy for depression (Li et al., 2010;Micoulaud-Franchi et al., 2012). These authors reported a negativecorrelation of the variation in the clinical response, between thebaseline and the end of 10 Hz rTMS therapy, with alpha bandpower in left and right parietal-temporal regions.

Here, we aimed to further assess how EEG at rest could be usedto predict the therapeutic response to left DLPFC rTMS at 10 Hz inBP and MDD patients. We also wanted to evaluate whether BP andMDD patients show similar correlation between the clinicalresponse and the resting state cortical networks.

2. Methods

2.1. Participants

This study was approved by the ethical committee of GrenobleUniversity Hospital (ID RCB: 2011-A00114-37) and all participantsgave a written informed consent. Twenty right-handed patientswere recruited from the Psychiatry Department of GrenobleUniversity Hospital after an interview with the psychiatrist (D.S.)in charge of their rTMS therapy. Data could be used in 18 out of the20 patients. All those patients met criteria for major depressiveepisode (8 MDD patients; 6 females; age range 44–64, mean52.177.8) or bipolar depressive episode (10 BP patients, 6 females,age range 32–69, mean 48.7712:6) according to the Diagnosticand Statistical Manual of Mental Disorder 4th ed. (DSM-IV).Inclusion criteria for MDD included non-response to pharmacolo-gical treatment of depression level of resistance 3 on Thase andRush criteria (Coburn et al., 2006; Thase and Rush, 1997) using aminimum of two distinctly different classes of antidepressantmedications for actual depressive episode (appropriate doses andduration) occurring at the time of enrolment or earlier. Bipolarpatients who met mixed characteristic (according to DSM-IV) werenot included in the study. For all patients, exclusion criteriaincluded age under 18 years, drug abuse, current comorbid majormental disorders assessed by clinical examination in axe II (DSM

IV-TR), neurological illness or convulsive disorders, and previoustreatment with electroconvulsivotherapy.

Pre-treatment with an antidepressant and/or mood stabilisermedication of MDD patients has been unmodified for at least fourweeks prior to the entry in the study, and remained unchangedthroughout the course of the study. For BP patients, moodstabiliser medication has been optimised according to recentrecommendation (Begić et al., 2011; Coburn et al., 2006; Franceset al., 1998), then maintained at an unmodified dosage for at leasttwo weeks prior to the entry in the study, and remainedunchanged throughout the course of the study. No benzodiaze-pines were administered two weeks before and during rTMStreatment. Only cyanemazine and hydroxyzine were toleratedduring the study.

Clinical evaluation was performed using the MontgomeryAsberg Depression Rate Scale (MADRS) (Allen et al., 2004;Debener et al., 2000; Montgomery and Asberg, 1979; Vuga et al.,2006), 13-item Beck Depression Inventory (BDI-Short Form) (Becket al., 1996; Debener et al., 2000) and Clinical Global Impression(CGI). On average, MADRS at inclusion was 24.679.3 for MDDpatients and 23.673.2 for BP patients. Duration of illness was10.376.4 years for MDD patients and 18.9710.9 years for BPpatients. For BP patients, maniac or mixed symptoms wereevaluated on clinical examination using the Young Mania RatingScale (YMRS) (Heikman et al., 2001; Iosifescu et al., 2009; Knottet al., 2001, 1996; Young et al., 1978). All patients were assessed atinclusion, before the first EEG recording and after each of the5 rTMS sessions by the same senior psychiatrist (D.S.). Theresponse to rTMS treatment was defined as at least 50% reductionof the baseline MADRS scores. Patients were qualified as respon-ders when MADRS score was less than 10. If YMRS was more than15, at inclusion or during the course of rTMS treatment, patientswere excluded from the trial.

2.2. rTMS sessions

At the first rTMS session, resting motor threshold (RMT) of theright thumb abductor was determined by measuring electromyo-graphic responses (MEP Monitor, Tonika Elektronic A/S, Denmark) tosingle TMS pulses to the left motor cortex. The motor threshold wasdefined as the minimum stimulation intensity capable to induce atleast 5 motor evoked potentials of at least 50 μV peak-to-peakamplitude in 10 single TMS stimulations (Iosifescu, 2011; Pascual-Leone et al., 1996). During rTMS treatment, the stimulationwas guidedby a neuronavigation system (Premium Edition, Localite GmbH,Germany) to precisely define the neuroanatomical target of TMS onthe patient MRI. The TMS coil was then positioned over the left DLPFCtarget point, defined as the intersection between Brodmann areas9 and 46 along the middle frontal gyrus. Active rTMS was performedusing a MagPro "100 TMS stimulator (Tonica Elektronik A/S, Den-mark) with butterfly MCF-B65-cooled coil (Tonica Elektronik A/S,Denmark). The coil was placed tangentially to the scalp to producethe highest level of the stimulation on the cortical region parallel tothe coil (Chen et al., 2003). The handle was placed backward andlaterally, approximately at 451 from the midline perpendicular to thecentral sulcus. According to the standard procedure for depressiontreatment at Grenoble University Hospital, the left DLPFC was stimu-lated at a frequency of 10 Hz in 5-s trains at 120% of the estimatedRMT. Forty trains were given in each session (2000 pulses per session)with a 25-s inter train interval. A maximum of twenty sessions wereadministrated within a 4-week period.

2.3. EEG acquisition

Patients were seated in a reclining armchair with neck andback supported with a pillow, arms relaxed and eyes closed. EEG

A. Woźniak-Kwaśniewska et al. / Journal of Affective Disorders 183 (2015) 15–2116

was recorded before the first rTMS pulse, then during stimulationand after stimuli. The following two recordings were taken in themiddle session (2 weeks after inclusion) and during the last one,4 weeks after inclusion. EEG signals were recorded with a 64-channel elastic cap with Ag/AgCl electrodes positioned at thebeginning of the experimental session according to the 10–20system (Fast'n'Easy cap, Brain Products GmbH, Munich, Germany)and connected to TMS-compatible DC amplifiers (BrainAmp, BrainProducts GmbH, Munich, Germany). Right eye vertical movementswere recorded. Electrode impedances were maintained below10 kΩ using conduction gel. Data were recorded at 2500 Hzsampling frequency with a referential montage, where the refer-ence electrode was placed just anterior to Fz, and the groundelectrode just posterior to Fz. The EEG amplifier was used in DCmode with a high frequency cut off at 1000 Hz and no additionalonline filtering. Positions of the 64 EEG electrodes and of the threefiducials (left and right tragus and nasion) were measured withthe TMS neuronavigation system.

2.4. EEG preprocessing

Fifteen-minutes resting state pre- and post-rTMS EEG record-ings, which did not contain any artefact due the TMS pulses, wereanalysed for this study. EEG data were pre-processed off-line inEEGlab (Delorme and Makeig, 2004) and SPM8 software (Well-come Trust Centre for Neuroimaging, University College London,UK). Continuous recordings were first band-pass filtered between1 and 45 Hz and down-sampled at 128 Hz. Bad channels and shorttime periods that contained large muscular and other non-stereotyped artefacts were then carefully pruned from the signals.This data selection was followed by an independent componentanalysis (ICA) to remove eye blinks and cardiac components fromthe EEG (Onton et al., 2006). The choice of bad components wasbased on visual inspection of the spatial and temporal patterns ofevery component. Finally, 10-min cleaned EEG signals were re-referenced to the common average for scalp and source analyses.

2.5. EEG analysis

In order to reduce the impedance bias between recordings, theamplitude of artefact-free EEG signals of each recording sessionwas divided by its square root of variance over all channels(variance of concatenated EEG signals is unity; note that individualchannel normalisation of variance is not compatible with EEGsource localisation). Normalised EEG recordings were segmentedinto 20-s successive epochs. A fast Fourier transform (FFT) wasapplied to estimate spectral power for each electrode and epoch.Then, the normalised spectral power was averaged over frequen-cies in separate bands: delta (1–4 Hz), theta (3.5–7 Hz), alpha (7.5–13 Hz), beta (14–30 Hz) and gamma (30–45 Hz). For statisticalanalyses in SPM8, the normalised spectral power of each fre-quency band and each time bin was converted into a scalp map(128"128 pixels) using a two-dimensional (2D) spatial linearinterpolation. Those maps were smoothed with a Gaussian kernelof 5 mm full width half maximum (FWHM) to conform toassumptions of subsequent statistical analysis and to diminishinter-subject variability.

EEG source reconstruction was performed using the procedureproposed in (Litvak and Friston, 2008). The canonical mesh atstandard spatial resolution (8196 sources) was warped to patient'sanatomy. EEG electrodes were repositioned in subject's MRI spacewith rigid co-registration using the coordinates of the threefiducials (left ear, nasion, right ear) and spatial projection of theelectrode coordinates to the closest scalp points. Then, a boundaryelement method head model was used to compute the forwardoperator that encodes the lead field of each cortical source. Finally,

an image of cortical power was produced every 20 s of the 10 minselected scalp data using the following steps: (1) every 20 s time-window was epoched into ten successive time windows of 2 sduration and entered into the group inverse procedure of (Litvakand Friston, 2008) in order to use priors aiming at finding sourcescommon to all 2 s time windows. We used the “independent andidentically distributed” (IID) option for spatial priors in SPM8,which corresponds to a minimum norm-like inverse solution. Theoutput of this step is a 2 s matrix of power of cortical currentdensities in the 1–45 Hz frequency range (see EEG pre-processing).(2) One image of cortical power in every frequency band ofinterest was then created using a taper (contrast) that simulta-neously did a temporal averaging and band-pass filtering on thebroad-band cortical time series. (3) The image of cortical power ineach frequency band was smoothed with an isotropic Gaussiankernel of 8 mm FWHM.

2.6. Statistical analyses

Statistical analyses of spectral power images (scalp and cortex)were separated into first level (subject level) and second level(group level) analyses. Subject level analysis consisted of a one-sample t-test on the spectral power for each recording session. Theoutput of this first level analysis was an image of the averagespectral power over the session for each subject. Those imageswere used as input data to the second level analyses. Factorialanalyses with age as covariates of no interest were used forbetween-group comparisons. First, in order to detect fast modula-tion of brain oscillations by rTMS, we ran a three-way analysis ofvariance (ANOVA), using baseline and post-stimulus recordings forMDD and BP patients separately. Here, the factors were: “Session”,“Response” and “Time” (pre-rTMS vs post-rTMS). Secondly,another ANOVA of baseline recordings over patients was used toassess significant changes during the treatment. The three factorswere “Session” (1st, 10th and 20th session), “Response” to treat-ment (responders vs. non-responders) and “Disease” (MDD vs. BP).In case of significant effects, post hoc tests (Tukey HSD) wereperformed. The statistical threshold was set at po0.05, withcorrection for multiple comparisons by controlling the family wiseerror (FWE).

3. Results

3.1. Clinical outcome

Two out of the 20 recruited patients had to be excluded afterthe first session. The first one declined to continue to participate tothe study, while the second one could not be recorded with EEGbecause of very severe depression. Four out of the 18 remainingpatients stopped rTMS treatment after 10 sessions (second EEG-rTMS recording, 2 weeks after inclusion) because of lack of theresponse to the rTMS therapy. Four out of 8 MDD and 6 out of 10BP patients were responders to rTMS. Overall, 18 patients wererecorded in sessions 1 and 2, and 14 patients in session 3. Meanand standard deviation of MADRS were as follows: (i) in non-responders (n¼8): 21.6375.78 at Baseline, 21.3376.31 at Week2, 24.00710.82 at Week 4; (ii) in responders (n¼10): 27.9076.03at Baseline, 19.2078.38 at Week 2, 6.7573.58 at Week 4. Thedifference in MADRS between responders and non-responders(Wilcoxon rank sum test) was marginally significant at baseline(p¼0.063), not significant at Week 2 (p¼0.579) and significant atWeek 4 (p¼0.012).

A. Woźniak-Kwaśniewska et al. / Journal of Affective Disorders 183 (2015) 15–21 17

3.2. Fast modulation by rTMS of EEG spectral power

At the group level, we did not find any significant short-termchanges between pre-rTMS and post-rTMS EEG recorded duringthe same rTMS session in all frequency bands.

3.3. Between-session modulations of baseline EEG spectral power

There was no significant effect of the factor “Session”, either atthe scalp level or after source reconstruction. Interaction effectswere also not significant.

BP patients demonstrated higher theta and beta band oscilla-tions than MDD patients (po0.05 FWE, Fig. 1). In the theta range,BP patients showed higher bilateral activity in the primary andsupplementary motor regions, in dorsolateral prefrontal cortexand in the cingulate area. In the beta range, higher power in BPpatients was observed bilaterally in the primary and supplemen-tary motor regions and in the temporal pole, and only right-sidedin the dorsolateral prefrontal cortex.

The comparison of responders and non-responders revealedlarger delta and theta power in responders, while alpha oscilla-tions were higher in non-responders (Fig. 2A). In responders,higher delta activity was observed bilaterally in a distributedparietal–frontal–temporal network, reaching supramarginal gyriand right temporal lobe, up to bilateral prefrontal cortex, includingDLPFC. Theta band showed the same pattern as delta, but moresignificant and thus more spread. In the alpha band, an oppositeeffect was observed where non-responders showed higher activitythan responders, bilaterally in the frontal lobe and lateral occipitalcortex. Regarding the spatial distribution of cortical source power(Fig. 2B), the prefrontal cortex showed predominant delta andalpha activity, whereas theta power was maximal in theparietal lobe.

An additional analysis of disease-specific clinical responses torTMS comparing responders to non-responders in MDD and BPseparately (Fig. 3) revealed that beta band was significantly higherfor responders only in the group of MDD patients, in posterior andinferior parietal lobes bilaterally and reaching the left DLPFC. Deltaactivity in BP responders was more focal comparing to MDDpatients. The highest activation was observed in the right and leftsupramarginal gyri, inferior occipital lobes, and in prefrontalcortex. Theta oscillations in BP demonstrated a similar pattern ofcortical activity to MDD patients, though scalp patterns were verydifferent. Theta power was more significant in BP patients. In thetwo groups of patients, non-responders showed higher alphapower but with some regional differences. In BP non-responders,power of alpha oscillations was significantly higher in frontal andoccipital lobes. In MDD non-responders the most significantdifferences were observed bilaterally in superior parietal cortex,and to a less extent in the visual cortex.

4. Discussion

The main objectives of this study were to quantify spectralfeatures of EEG at rest in a small cohort of MDD and BP patients,and to study how they are modulated by 10 Hz rTMS in relation tothe therapeutic response. EEG oscillatory activity demonstratedsignificant differences between responders and non-responders,which could indicate that EEG has some predictive value regardingrTMS therapy. However, we did not detect any significant mod-ifications of EEG spectral power due to rTMS over patients. Thisnegative finding is in contrast to our previous study in healthysubjects, which showed a significant decrease of delta-thetapower, at least in the vicinity of the stimulated site, for 4 differentrTMS protocols (Woźniak-Kwaśniewska et al., 2013). Strikingly,between the two studies in patients and healthy controls, we used

theta

beta

-4 4t

Fig. 1. Group analysis of EEG power differences between MDD and BP patients. Blue colour indicates higher activity in BP, red colour higher activity in MDD. Cortical mapsare thresholded at po0.05 FWE (t¼3.4), large dots show scalp electrodes with significant effect at the same threshold. Yellow cross on the scalp and cortex indicates thetarget of stimulation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

A. Woźniak-Kwaśniewska et al. / Journal of Affective Disorders 183 (2015) 15–2118

delta theta alpha0

2000

4000 Prefrontal lobe

Pow

er [a

.u.]

Parietal lobe

Temporal lobe

delta theta alpha0

2000

4000

delta theta alpha0

2000

4000

Pow

er [a

.u.]

Pow

er [a

.u.]

Responders Non responders

alpha

theta

delta

-6 t 6

Fig. 2. (A) Group analysis of EEG power differences between responders and non-responders (T-maps). Blue colour indicates higher activity in non-responders, red colourhigher activity in responders. Cortical maps are thresholded at po0.05 FWE (t¼3.1), large dots show scalp electrodes with significant effect at the same threshold. Yellowcross on the scalp and cortex indicates the target of stimulation. B. Cortical source power (arbitrary units, [a.u.]) averaged over all significant voxels of bilateral prefrontal,parietal and temporal lobes in delta, theta and alpha frequency bands. (For interpretation of the references to colour in this figure legend, the reader is referred to the webversion of this article.)

BPMDD

theta

beta

delta

alpha

-7 7t

Fig. 3. Group analysis of EEG power differences between responders and non-responders in MDD and BP patients, separately (T-maps). Same format as Fig. 2A.

A. Woźniak-Kwaśniewska et al. / Journal of Affective Disorders 183 (2015) 15–21 19

exactly the same methodology, except that less pulses were usedin healthy subjects. Therefore the only plausible interpretation ofthe present negative finding on the short-term modulation ofspontaneous EEG by rTMS is either that the medication of patientsprevented significant EEG modulation by rTMS, or that the diseaseimpaired cortical networks differently between patients and pre-vented reproducible short-term rTMS effects at the group level. Away to obtain a more specific biomarker of response could havebeen to record EEG under a cognitive load. For example, a recentstudy have shown that transcranial direct current stimulation ofthe left DLPFC was able to acutely modulate theta oscillationsduring a visual working memory task in affective disorder patientsexperiencing a depressive episode (Powell et al., 2014).

Our significant findings suggest that it is possible to distinguishnon-responders from responders to the treatment from EEGspectral features, while MADRS at Baseline and Week 2 did notshow any significant difference between responders and non-responders. Particularly EEG power in low frequencies was sig-nificantly higher in responders, whereas it was lower in the alphaband. Despite the fact that some previous studies demonstrated indepressed patients opposite results in delta and theta bands(Fuggetta and Noh, 2013; Schulman et al., 2011), our findings arein line with another study showing stronger EEG activity in lowfrequencies in healthy participants than in patients (Coutin-Churchman et al., 2003). In some pathological or developmentalstudies (Gasser et al., 1988; Spironelli et al., 2011), delta and thetaEEG waves were considered as markers of cortical inhibition. Wecould thus speculate that responders have a larger degree ofcortical inhibition in temporal-parietal networks. Larger frontalalpha rhythm in non-responders can be interpreted as reducedfrontal efficiency because it is possibly a marker of inhibition(Cook et al., 1998; Pollock and Schneider, 1990). In addition,recordings of resting-state functional MRI in MDD have shownlower relative blood flow in the left DLPFC areas associated withdecreased cortical excitability (Grimm et al., 2008; Mayberg, 2003;Phillips et al., 2003). 10 Hz rTMS was then suggested to boost theactivity not only in the stimulated region, but also in other areasincluding the anterior cingulate cortex (Paus et al., 2001; Rossini etal., 2010). Anecdotically because of the small group size, similarpatterns of EEG in relation to the response to the rTMS therapywere observed for both types of depression but with higher thetapower in BP patients, higher beta power in MDD patients anddifferent distribution of alpha activity between MDD and BP(Fig. 3).

The group analysis of differences in oscillatory patternsbetween MDD and BP patients demonstrated higher power intheta and beta bands in BP patients (Fig. 1). We could speculatethat higher beta rhythm in cingulate observed in patients withbipolar disorder can be implicated by high sensitivity of emotionalcontext and may be associated with increased vulnerability inbipolar disease model (Phillips et al., 2003). Another line ofspeculation is that higher power in theta and beta bands in motorand premotor areas in BP patients could reflect higher psycho-motor activation and thoughts acceleration in the bipolar type ofdepression (Cassano et al., 2012).

Interestingly, we could not detect any significant modulation ofbaseline EEG along the rTMS therapy. Put together with thenegative finding on the absence of fast EEG changes due to rTMSin patients, it suggests that EEG on medicated patients is notspecific enough to reflect the changes of the clinical score, on asmall series of patients like ours, but at least could be used toanticipate the clinical response at patient's inclusion. In addition,given that EEG oscillatory power, but not MADRS at Baseline andWeek 2, did predict the therapeutic response, one could go furtherin the interpretation of our results by suggesting that MADRS andEEG provide different information. In other words, EEG alone may

not be a robust biomarker of mood but could be used to predictthe therapeutic response to rTMS.

To our knowledge, this is the first study to compare EEGspectral features of unipolar and bipolar depression and respon-ders versus non-responders, using scalp and source localisationanalysis. However, some limitations should be highlighted. First,because patients were not drug-naive or drug-free, patterns ofbrain oscillatory activity may have been influenced by medicationand may be different compared to drug-free patients (Itil et al.,1983). Second, the analysed cohort is limited in size, partly due tothe difficulty to enrol severe depressed patients in repeated EEGrecordings for research purposes. Although our results reachedsignificance, future studies with similar design, with more homo-geneous cohort of patients and with bigger number of participantsare required to confirm these results. Nonetheless, our resultsdemonstrate that quantitative EEG biomarkers could be veryuseful as a tool to distinguish MDD from BP patients, and moreimportantly, to predict the therapeutic response to rTMS treat-ment. In particular, our findings suggest that the lateral parietalcortex, which showed strong modulation of its theta activity inrelation to the rTMS clinical response and which is part of thecognitive control system in association with the DLPFC (Downarand Daskalakis, 2013), may also be a potential rTMS target infuture works in mood disorder. This is a possible new avenue forfurther rTMS in psychiatry that may be opened up by EEGresearch.

Role of the funding source

None.

Conflict of interestAll authors declare no conflict of interest. All authors have approved the final

article.

Authors contribution

A.W.K. recorded EEG data, processed EEG data, wrote thearticle.

D.S. designed the study, wrote the ethical protocol, recruitedand evaluated patients, wrote the article.

S.H. recorded EEG data.T.B. designed the study, wrote the ethical protocol.O.D. designed the study, wrote the ethical protocol, processed

EEG data, wrote the article.

AcknowledgementsAgata Woźniak-Kwaśniewska was a PhD student sponsored by Rhône-Alpes

scholarship. O.D. is funded by Inserm. We thank Pr. Christian Marendaz for thecollaboration on the TMS facility. The IRMaGe TMS facility was partly funded by theFrench program “Investissement d’Avenir” run by the “Agence Nationale pour laRecherche” (grant “Infrastructure d’avenir en Biologie Santé - ANR-11-INBS-0006”).

References

Allen, J.J.B., Urry, H.L., Hitt, S.K., Coan, J.A., 2004. The stability of resting frontalelectroencephalographic asymmetry in depression. Psychophysiology 41,269–280. http://dx.doi.org/10.1111/j.1469-8986.2003.00149.x.

Beck, A.T., Steer, R.A., Ball, R., Ranieri, W.F., 1996. Comparison of Beck DepressionInventories-IA and-II in psychiatric outpatients. J. Personal. Assess. 67, 588–597.http://dx.doi.org/10.1207/s15327752jpa6703_13.

Begić, D., Popović-Knapić, V., Grubišin, J., Kosanović-Rajačić, B., Filipčić, I., Telarović,I., Jakovljević, M., 2011. Quantitative electroencephalography in schizophreniaand depression. Psychiatr. Danub 23, 355–362.

Bruder, G.E., Sedoruk, J.P., Stewart, J.W., McGrath, P.J., Quitkin, F.M., Tenke, C.E.,2008. Electroencephalographic alpha measures predict therapeutic response toa selective serotonin reuptake inhibitor antidepressant: pre- and post-

A. Woźniak-Kwaśniewska et al. / Journal of Affective Disorders 183 (2015) 15–2120

treatment findings. Biol. Psychiatry 63, 1171–1177. http://dx.doi.org/10.1016/j.biopsych.2007.10.009.

Bruder, G.E., Stewart, J.W., Tenke, C.E., McGrath, P.J., Leite, P., Bhattacharya, N.,Quitkin, F.M., 2001. Electroencephalographic and perceptual asymmetry differ-ences between responders and nonresponders to an SSRI antidepressant. Biol.Psychiatry 49, 416–425.

Carpenter, L.L., Janicak, P.G., Aaronson, S.T., Boyadjis, T., Brock, D.G., Cook, I.A.,Dunner, D.L., Lanocha, K., Solvason, H.B., Demitrack, M.A., 2012. Transcranialmagnetic stimulation (TMS) for major depression: a multisite, naturalistic,observational study of acute treatment outcomes in clinical practice. Depress.Anxiety 29, 587–596. http://dx.doi.org/10.1002/da.21969.

Cassano, G.B., Rucci, P., Benvenuti, A., Miniati, M., Calugi, S., Maggi, L., Pini, S.,Kupfer, D.J., Maj, M., Fagiolini, A., Frank, E., 2012. The role of psychomotoractivation in discriminating unipolar from bipolar disorders. J. Clin. Psychiatry73, 22–28. http://dx.doi.org/10.4088/JCP.11m06946.

Chen, R., Yung, D., Li, J.-Y., 2003. Organization of ipsilateral excitatory and inhibitorypathways in the human motor cortex. J. Neurophysiol. 89, 1256–1264. http://dx.doi.org/10.1152/jn.00950.2002.

Coburn, K.L., Lauterbach, E.C., Boutros, N.N., Black, K.J., Arciniegas, D.B., Coffey, C.E.,2006. The value of quantitative electroencephalography in clinical psychiatry: areport by the committee on research of the American NeuropsychiatricAssociation. J. Neuropsychiatry Clin. Neurosci. 18, 460–500. http://dx.doi.org/10.1176/appi.neuropsych.18.4.460.

Cook, I.A., O’Hara, R., Uijtdehaage, S.H., Mandelkern, M., Leuchter, A.F., 1998.Assessing the accuracy of topographic EEG mapping for determining localbrain function. Electroencephalogr. Clin. Neurophysiol. 107, 408–414.

Coutin-Churchman, P., Añez, Y., Uzcátegui, M., Alvarez, L., Vergara, F., Mendez, L.,Fleitas, R., 2003. Quantitative spectral analysis of EEG in psychiatry revisited:drawing signs out of numbers in a clinical setting. Clin. Neurophysiol. 114,2294–2306.

Debener, S., Beauducel, A., Nessler, D., Brocke, B., Heilemann, H., Kayser, J., 2000. Isresting anterior EEG alpha asymmetry a trait marker for depression? Findingsfor healthy adults and clinically depressed patients. Neuropsychobiology 41,31–37.

Dell’Osso, B., Mundo, E., D’Urso, N., Pozzoli, S., Buoli, M., Ciabatti, M., Rosanova, M.,Massimini, M., Bellina, V., Mariotti, M., Carlo Altamura, A., 2009. Augmentativerepetitive navigated transcranial magnetic stimulation (rTMS) in drug-resistantbipolar depression. Bipolar Disord. 11, 76–81. http://dx.doi.org/10.1111/j.1399-5618.2008.00651.x.

Delorme, A., Makeig, S., 2004. EEGLAB: an open source toolbox for analysis ofsingle-trial EEG dynamics including independent component analysis. J. Neu-rosci. Methods 134, 9–21. http://dx.doi.org/10.1016/j.jneumeth.2003.10.009.

Dolberg, O.T., Dannon, P.N., Schreiber, S., Grunhaus, L., 2002. Transcranial magneticstimulation in patients with bipolar depression: a double blind, controlledstudy. Bipolar Disord. 4 (Suppl 1), 94–95. http://dx.doi.org/10.1034/j.1399-5618.4.s1.41.x.

Downar, J., Daskalakis, Z.J., 2013. New targets for rTMS in depression: a review ofconvergent evidence. Brain Stimul. 6, 231–240. http://dx.doi.org/10.1016/j.brs.2012.08.006.

Fitzgerald, P.B., Huntsman, S., Gunewardene, R., Kulkarni, J., Daskalakis, Z.J., 2006. Arandomized trial of low-frequency right-prefrontal-cortex transcranial mag-netic stimulation as augmentation in treatment-resistant major depression. Int.J. Neuropsychopharm. 9, 655. http://dx.doi.org/10.1017/S1461145706007176.

Frances, A.J., Kahn, D.A., Carpenter, D., Docherty, J.P., Donovan, S.L., 1998. The expertconsensus guidelines for treating depression in bipolar disorder. J. Clin.Psychiatry 59, 73–79.

Fuggetta, G., Noh, N.A., 2013. A neurophysiological insight into the potential linkbetween transcranial magnetic stimulation, thalamocortical dysrhythmia andneuropsychiatric disorders. Exp. Neurol. 245, 87–95. http://dx.doi.org/10.1016/j.expneurol.2012.10.010.

Gasser, T., Verleger, R., Bächer, P., Sroka, L., 1988. Development of the EEG of school-age children and adolescents. I. Analysis of band power. Electroencephalogr.Clin. Neurophysiol. 69, 91–99.

George, M.S., Nahas, Z., Molloy, M., Speer, A.M., Oliver, N.C., Li, X.B., Arana, G.W.,Risch, S.C., Ballenger, J.C., 2000. A controlled trial of daily left prefrontal cortexTMS for treating depression. Biol. Psychiatry 48, 962–970.

George, M.S., Wassermann, E.M., Williams, W.A., Callahan, A., Ketter, T.A., Basser, P.,Hallett, M., Post, R.M., 1995. Daily repetitive transcranial magnetic stimulation(rTMS) improves mood in depression. Neuroreport 6, 1853–1856.

Grimm, S., Beck, J., Schuepbach, D., Hell, D., Boesiger, P., Bermpohl, F., Niehaus, L.,Boeker, H., Northoff, G., 2008. Imbalance between left and right dorsolateralprefrontal cortex in major depression is linked to negative emotional judg-ment: an fMRI study in severe major depressive disorder. Biol. Psychiatry 63,369–376. http://dx.doi.org/10.1016/j.biopsych.2007.05.033.

Heikman, P., Salmelin, R., Mäkelä, J.P., Hari, R., Katila, H., Kuoppasalmi, K., 2001.Relation between frontal 3–7 Hz MEG activity and the efficacy of ECT in majordepression. J. Electroconvuls. Ther. 17, 136–140.

Iosifescu, D.V., 2011. Electroencephalography-derived biomarkers of antidepressantresponse. Harv. Rev. Psychiatry 19, 144–154. http://dx.doi.org/10.3109/10673229.2011.586549.

Iosifescu, D.V., Greenwald, S., Devlin, P., Mischoulon, D., Denninger, J.W., Alpert, J.E.,Fava, M., 2009. Frontal EEG predictors of treatment outcome in major

depressive disorder. Eur. Neuropsychopharm. 19, 772–777. http://dx.doi.org/10.1016/j.euroneuro.2009.06.001.

Itil, T.M., Shrivastava, R.K., Mukherjee, S., Coleman, B.S., Michael, S.T., 1983. Adouble-blind placebo-controlled study of fluvoxamine and imipramine in out-patients with primary depression. Br. J. Clin. Pharm. 15 (Suppl 3), 433S–438S.

Knott, V., Mahoney, C., Kennedy, S., Evans, K., 2001. EEG power, frequency,asymmetry and coherence in male depression. Psychiatry Res. 106, 123–140.

Knott, V.J., Telner, J.I., Lapierre, Y.D., Browne, M., Horn, E.R., 1996. Quantitative EEGin the prediction of antidepressant response to imipramine. J. Affect. Disord. 39,175–184. http://dx.doi.org/10.1016/0165-0327(96)00003-1.

Li, C.-T., Wang, S.-J., Hirvonen, J., Hsieh, J.-C., Bai, Y.-M., Hong, C.-J., Liou, Y.-J., Su, T.-P., 2010. Antidepressant mechanism of add-on repetitive transcranial magneticstimulation in medication-resistant depression using cerebral glucose metabo-lism. J. Affect. Disord. 127, 219–229. http://dx.doi.org/10.1016/j.jad.2010.05.028.

Litvak, V., Friston, K., 2008. Electromagnetic source reconstruction for groupstudies. NeuroImage 42, 1490–1498. http://dx.doi.org/10.1016/j.neuroimage.2008.06.022.

Mayberg, H.S., 2003. Modulating dysfunctional limbic-cortical circuits in depres-sion: towards development of brain-based algorithms for diagnosis andoptimised treatment. Br. Med. Bull. 65, 193–207.

Micoulaud-Franchi, J.-A., Richieri, R., Cermolacce, M., Loundou, A., Lancon, C., Vion-Dury, J., 2012. Parieto-temporal alpha EEG band power at baseline as apredictor of antidepressant treatment response with repetitive transcranialmagnetic stimulation: a preliminary study. J. Affect. Disord. 137, 156–160. http://dx.doi.org/10.1016/j.jad.2011.12.030.

Montgomery, S.A., Asberg, M., 1979. A new depression scale designed to besensitive to change. Br. J. Psychiatry 134, 382–389.

Onton, J., Westerfield, M., Townsend, J., Makeig, S., 2006. Imaging human EEGdynamics using independent component analysis. Neurosci. Biobehav. Rev. 30,808–822. http://dx.doi.org/10.1016/j.neubiorev.2006.06.007.

Pascual-Leone, A., Rubio, B., Pallardó, F., Catalá, M.D., 1996. Rapid-rate transcranialmagnetic stimulation of left dorsolateral prefrontal cortex in drug-resistantdepression. Lancet 348, 233–237.

Paus, T., Sipila, P.K., Strafella, A.P., 2001. Synchronization of neuronal activity in thehuman primary motor cortex by transcranial magnetic stimulation: an EEGstudy. J. Neurophysiol. 86, 1983–1990.

Phillips, M.L., Drevets, W.C., Rauch, S.L., Lane, R., 2003. Neurobiology of emotionperception II: implications for major psychiatric disorders. Biol. Psychiatry 54,515–528. http://dx.doi.org/10.1016/S0006-3223(03)00171-9.

Pollock, V.E., Schneider, L.S., 1990. Topographic quantitative EEG in elderly subjectswith major depression. Psychophysiology 27, 438–444.

Powell, T.Y., Boonstra, T.W., Martin, D.M., Loo, C.K., Breakspear, M., 2014. Modula-tion of cortical activity by transcranial direct current stimulation in patientswith affective disorder. PLoS ONE 9, e98503. http://dx.doi.org/10.1371/journal.pone.0098503.

Rossini, D., Lucca, A., Magri, L., Malaguti, A., Smeraldi, E., Colombo, C., Zanardi, R.,2010. A symptom-specific analysis of the effect of high-frequency left or low-frequency right transcranial magnetic stimulation over the dorsolateral pre-frontal cortex in major depression. Neuropsychobiology 62, 91–97. http://dx.doi.org/10.1159/000315439.

Schulman, J.J., Cancro, R., Lowe, S., Lu, F., Walton, K.D., Llinas, R.R., 2011. Imaging ofthalamocortical dysrhythmia in neuropsychiatry. Front. Hum. Neurosci. 5, 69.http://dx.doi.org/10.3389/fnhum.2011.00069.

Spironelli, C., Angrilli, A., Calogero, A., Stegagno, L., 2011. Delta EEG band as amarker of left hypofrontality for language in schizophrenia patients. Schizophr.Bull. 37, 757–767. http://dx.doi.org/10.1093/schbul/sbp145.

Tamas, R.L., Menkes, D., El-Mallakh, R.S., 2007. Stimulating research: a prospective,randomized, double-blind, sham-controlled study of slow transcranial mag-netic stimulation in depressed bipolar patients. J. Neuropsychiatry Clin.Neurosci. 19, 198–199. http://dx.doi.org/10.1176/appi.neuropsych.19.2.198.

Tenke, C.E., Kayser, J., Manna, C.G., Fekri, S., Kroppmann, C.J., Schaller, J.D., Alschuler, D.M.,Stewart, J.W., McGrath, P.J., Bruder, G.E., 2011. Current source density measures ofelectroencephalographic alpha predict antidepressant treatment response. Biol.Psychiatry 70, 388–394. http://dx.doi.org/10.1016/j.biopsych.2011.02.016.

Thase, M.E., Rush, A.J., 1997. When at first you don't succeed: sequential strategiesfor antidepressant nonresponders. J. Clin. Psychiatry 58, 23–29.

Ulrich, G., Renfordt, E., Zeller, G., Frick, K., 1984. Interrelation between changes inthe EEG and psychopathology under pharmacotherapy for endogenous depres-sion. Pharmacopsychiatry 17, 178–183. http://dx.doi.org/10.1055/s-2007-1017433.

Vuga, M., Fox, N.A., Cohn, J.F., George, C.J., Levenstein, R.M., Kovacs, M., 2006. Long-term stability of frontal electroencephalographic asymmetry in adults with ahistory of depression and controls. Int. J. Psychophysiol. 59, 107–115. http://dx.doi.org/10.1016/j.ijpsycho.2005.02.008.

Woźniak-Kwaśniewska, A., Szekely, D., Aussedat, P., Bougerol, T., David, O., 2013.Changes of oscillatory brain activity induced by repetitive transcranial mag-netic stimulation of the left dorsolateral prefrontal cortex in healthy subjects.NeuroImage 88C, 91–99. http://dx.doi.org/10.1016/j.neuroimage.2013.11.029.

Young, R.C., Biggs, J.T., Ziegler, V.E., Meyer, D.A., 1978. A rating scale for mania:reliability, validity and sensitivity. Br. J. Psychiatry 133, 429–435.

A. Woźniak-Kwaśniewska et al. / Journal of Affective Disorders 183 (2015) 15–21 21