DSCTa

BsPd

MRat

Rtah

Ca

K

Emee(anumatEois

b(

F

A

R

0d

ifferential Effects of High-Dose Magneticeizure Therapy and Electroconvulsive Shock onognitive Function

imothy Spellman, Shawn M. McClintock, Herbert Terrace, Bruce Luber, Mustafa M. Husain,nd Sarah H. Lisanby

ackground: Magnetic seizure therapy (MST) is under investigation as an alternative form of convulsive therapy that induces more focaleizures and spares cortical regions involved in memory. With a newly expanded version of the Columbia University Primate Cognitiverofile, we compared the cognitive effects of high-dose MST delivered at 100 Hz (6 � seizure threshold) with electroconvulsive shock (ECS)elivered at 2.5 � seizure threshold.

ethods: Daily high-dose MST, ECS, and sham (anesthesia-only) were administered for 4 weeks each in a within-subject crossover design.hesus macaques (n � 3) were trained on five cognitive tasks assessing automatic memory, anterograde learning and memory, combinednterograde and retrograde simultaneous chaining, and spatial and serial working memory. Acutely after each intervention, monkeys wereested on the cognitive battery twice daily, separated by a 3-hour retention interval.

esults: Subjects were slower to complete criterion tasks (p values � .0001) after ECS, compared with sham and high-dose MST. Moreover,ime to task-completion after high-dose MST did not differ from sham. Of six measures of accuracy, treatment effects were found in four; inll of these, ECS but not MST fared worse than sham. On all accuracy and time to completion measurements, subjects performed as well afterigh-dose MST as subjects from a previous study on moderate-dose MST.

onclusions: These findings provide evidence that high-dose MST results in benign acute cognitive side-effect profile relative to ECS and

re in line with our previous studies.

ey Words: Cognitive, ECT, impairment, magnetic, seizure, TMS

lectroconvulsive therapy (ECT) is a highly effective treat-ment for major affective disorders (1). Although its sideeffect profile has been substantially improved through

odifications in electrode placement, stimulus dosage, andlectrical parameters, ECT carries a risk of neurocognitive sideffects (2,3). The most prominent of these are retrograde amnesiaRA), which impairs retrieval of events before treatment, andnterograde amnesia (AA), which impairs the ability to encodeew events into memory. Magnetic seizure therapy (MST) isnder development as an alternative convulsive technique toinimize the neurocognitive adverse effects of convulsive ther-

py while maintaining antidepressant efficacy (4,5). Preclinicalesting suggests that MST might have several advantages overCT, including increased precision of stimulation, greater controlf intracerebral spatial distribution, lack of susceptibility tompedance from surface tissues, and the sparing of deep braintructures (6,7).

The effects of MST on neurocognitive functioning haveeen examined in monkeys (8) and subsequently in humans9). In a within-subject study design, two monkeys received

rom the Division of Brain Stimulation and Therapeutic Modulation (TS,SMM, HT, BL, SHL), Department of Psychiatry, Columbia University / NewYork State Psychiatric Institute; Department of Psychology (HT), Colum-bia University, New York, New York; and the Neurostimulation ResearchLaboratory (SMM, MMH), University of Texas Southwestern Medical Cen-ter, Department of Psychiatry, Dallas, Texas.

ddress reprint requests to Sarah H. Lisanby, M.D., Division of Brain Stimu-lation and Therapeutic Modulation, Department of Psychiatry, ColumbiaUniversity/ New York State Psychiatric Institute, 1051 Riverside Drive,Unit 21, New York, NY 10032; E-mail: slisanby@columbia.edu.

eceived September 19, 2007; revised November 13, 2007; accepted No-

vember 29, 2007.

006-3223/08/$34.00oi:10.1016/j.biopsych.2007.11.024

MST, electroconvulsive shock (ECS), and sham treatment(anesthesia only) (8). Subjects were tested with a battery ofcognitive tasks previously shown to be sensitive to the effectsof ECS (10). Results showed that moderate-dose MST, admin-istered at 2.5 � the seizure threshold, resulted in fewercognitive adverse effects than ECS, also provided at 2.5 �seizure threshold. Specifically, subjects took significantly lesstime to complete cognitive tasks and showed greater accuracyafter moderate-dose MST than ECS.

Similar results were found in depressed patients who receivedMST (9). In a within-subject design, 10 patients received two MSTtreatments (one at seizure threshold and the other at maximalstimulator output that was 1.5–2 � seizure threshold) during anacute course of ECT. Relative to performance on neurocognitivemeasures after ECT, patients performed significantly better onmeasures of category fluency, sentence recognition, face recog-nition, and visual cancellation after MST treatments. Moreover,patients showed faster orientation recovery after MST than afterECT (9). Longer reorientation time has been associated withlonger-term retrograde amnesia (11).

Recently, we analyzed neurocognitive effects of a completeacute course of MST in 20 depressed patients in a double-masked, randomized, controlled clinical trial comparing twoforms of MST (both 1.5–2 � seizure threshold) (12,13). After anaverage of 9.0 (� 2.8) MST treatments, no change was seen inglobal cognitive function or verbal and visual anterograde mem-ory. Subjective improvement in cognitive functioning was seenon the Squire Self-Rating Scale of Memory Function and theCognitive Failures Questionnaire. However, regarding retrogradeamnesia, performance on the Autobiographical Memory Inter-view decreased minimally, and there was no difference on theGoldberg Remote Memory Questionnaire (14). Both this and theprior human studies used doses of MST that were close to or

slightly above seizure threshold.

BIOL PSYCHIATRY 2008;63:1163–1170© 2008 Society of Biological Psychiatry

clooct(aTlasoh

1dptdsa4�cd

Fts

1164 BIOL PSYCHIATRY 2008;63:1163–1170 T. Spellman et al.

w

Although monkey and human studies have found the neuro-ognitive effects of MST to be benign relative to ECT, deviceimitations constrained those studies to low and moderate dosesf MST delivered at 50 Hz. However, the efficacy and side effectsf ECT are known to be highly dosage-sensitive. This is espe-ially true for right unilateral (RUL) ECT. The RUL ECT adminis-ered at seizure threshold has a weak antidepressant effect15,16), but when the dose is increased, it is more efficacious (17)nd might also result in greater cognitive impairment (18–20).he viability of MST as a therapeutic alternative to ECT depends

argely on whether it can retain cognitive advantages whendministered at doses potent enough to approach the antidepres-ant effect of ECT. Thus, characterization of the cognitive effectsf MST at higher doses is warranted to support subsequentuman work.

Recently a more powerful MST device capable of sustaining00% of maximal stimulator output at 100 Hz for trains of 10-securation (1000 pulses/train) became available (21). Here weresent the first evaluation of the cognitive effects of chronicreatment with 100-Hz MST. A dose of 6 � seizure thresholduration was selected for MST stimulation in order to paralleltudies that have found antidepressant efficacy with RUL ECT atdose of 6 � seizure threshold (17). Rhesus monkeys receivedweeks of daily high-dose MST (6 � seizure threshold), ECS (2.5seizure threshold), and sham (anesthesia alone). We also

ompared high-dose MST to a prior experiment with moderate-ose MST in the same subjects to evaluate MST dose effects.

igure 2. Task 2: New target learning. The subject must learn, by selecting she target and then correctly select the target on four of five consecutive tr

ame between trials on a given day, but it changes between days.

ww.sobp.org/journal

Further, we present an expanded cognitive battery for nonhu-man primates to assess the impact of convulsive therapy onspatial working memory and serial probe recognition (SPR).

Methods and Materials

SubjectsThis study was approved by the Institutional Animal Care and

Use Committee (IACUC) of the New York State PsychiatricInstitute and Columbia University. The subjects were threepathogen-free male rhesus macaca mulatta monkeys obtainedfrom the same National Institutes of Health (NIH) breedingcolony. Mean age upon entering the study was 83 (� 26) months,mean weight was 8 (� 1) kg, and all three were past sexualmaturity. The approximate age equivalent in human years was20.8 (� 6.5) years (22,23). They were individually housed in acolony where they maintained a 12-hour light-dark cycle. Sub-jects had ad libitum access to water and received daily feedingsof standard monkey chow (LabDiet, W.F. Fischer & Son, Somer-ville, New Jersey), fruit, food pellets given as positive reinforce-ment during cognitive testing, and treats hidden in enrichmenttoys.

Cognitive TestingDetails of the cognitive testing apparatus, stimuli presenta-

tion, and training procedures have been reported elsewhere(8,10). We used the Columbia University Primate Cognitive

i through trial and error, which stimulus among a field of eight distracters isistracters change between trials and days. The target stimulus remains the

Figure 1. Task 1: Recall of an over-learned stimulus.The subject must correctly select which among a fieldof 16 stimuli is the target. The target remains the sameon every trial of every day, whereas the distracters andtheir screen positions change randomly between tri-als and days. A food pellet is given for each correcttrial. Trials are presented until the subject gets four offive consecutive trials correct.

timulials. D

Paopossdtm(esscmssmsghw(aaww

ant

F y triap day o

T. Spellman et al. BIOL PSYCHIATRY 2008;63:1163–1170 1165

rofile (CUPCP), which consists of three cognitive tasks thatssess learning and memory. Tasks consist of stimuli presentedn touchscreen monitors; stimuli consist of randomly chosenictures, which the subject selects by touching. Task 1, anrientation task (long-term or automatic memory), requires theubject to select a single target stimulus from a field of distractertimuli; this target stimulus does not change between trials orays. This task was given in the morning, immediately afterreatment. The subject was required to get 4 of 5 trials correct toove on to Task 2 (Figure 1). Task 2, a variable target task

anterograde learning and memory), required the subject to learnach day, by trial and error, which stimulus among a field of 9timuli was the target. The distracter stimuli but not the targettimulus changed from trial to trial; the subject was required toorrectly select the target in 4 of 5 consecutive trials in order toove on to Task 3 (Figure 2). Task 2 was given in the morning

ession immediately after treatment and again in the afternoonession, 3 hours after treatment. Task 3, a serial learning andemory task (anterograde and retrograde memory), required the

ubject to learn the correct sequence in which to respond to aroup of 3 stimuli. This task consisted of both new lists, whichad to be learned each day through trial and error, and old lists,hich had been learned between 1 and 3 years previously

Figure 3). New lists were given in both the morning andfternoon sessions, whereas old lists were given only in thefternoon. For the present study, two new tasks assessingorking memory were added to the CUPCP: Task 4, spatialorking memory; and Task 5, SPR.

Task 4 —Spatial Working Memory. This task was modeledfter a virtual radial arm maze task in which a subject mustavigate through a maze to different locations, without returningo the same location twice (24). Rather than moving through

igure 3. Task 3: Sequence learning and recall. The subject must learn, bresented. This task consisted of new lists, with a novel list presented each

three-dimensional space, the subject was required to move thecursor to different points on a two-dimensional touchscreenmonitor. A list of four, five, or six targets was arranged in a spatialconfiguration that changed randomly from trial to trial. All stimuliin a given trial were identical, so that only their position on thescreen differentiated them. In a given trial, subjects were re-quired to select each stimulus exactly once, in whatever se-quence they preferred. All stimuli remained on the screen untilthe end of the trial, and returning to a previously-selectedstimulus would result in an incorrect trial (Figure 4). Subjectsperformed this task during the afternoon session.

Task 5—SPR. The SPR task was designed to test workingmemory and was modeled after the Sternberg working memorytask (25). In this task, a list of four, five, or six distinct items waspresented, one item at a time (Figure 5). After list presentationand a 2-sec delay, a sample stimulus was presented and subjectsindicated, by responding to “yes” and “no” icons, whether or notthe target had been in the list. This task was also administeredduring the afternoon session.

Study DesignThe study used a within-subject design, allowing each mon-

key to serve as its own control subject. The final 2 weeks of datacollection before each subject’s first treatment were designatedas the baseline period. In randomized order, subjects received 20days of ECS, high-dose MST, and sham treatment (anesthesiaalone). A recovery phase, with a minimum of 20 days and amaximum of 30 days, immediately followed each treatmentphase. The length of the recovery period was determined by areturn to baseline performance. Titration for seizure thresholdwas performed on days 1 and 11 of both MST and ECS treatmentphases. Treatment order was counterbalanced between subjects:

l and error, the correct sequence in which to select three simultaneouslyn treatment days, and old lists, learned between 1 and 3 years previously.

Figure 4. Task 4: Spatial working memory task. Four, five, orsix identical stimuli are presented on the screen simulta-neously. The subject must select each stimulus in turn with-out repeating a selection. Because the stimuli are identicaland remain on the screen until the trial is over, they can bedifferentiated only by their respective locations on thescreen. Between selections on a given trial, the subjectmust return to the center and select a “reset” stimulus.

www.sobp.org/journal

sMt

E

arhwdwMrbsbSs(hfpww5sHi

Fpgaw

1166 BIOL PSYCHIATRY 2008;63:1163–1170 T. Spellman et al.

w

ubject 1 received ECS, then MST, then sham; subject 2 receivedST, then ECS, then sham; subject 3 received sham, then ECS,

hen MST.

CS, MST, and Sham InterventionsDetails of the interventions, including seizure threshold titration,

nesthesia, seizure monitoring, and vital sign monitoring have beeneported previously (8,10). Briefly, anesthesia consisted of metho-exital (.5 mg/kg IV) and succinylcholine (2.5 mg/kg IV). The MSTas administered via a custom, modified, 100-Hz Magstim MSTevice (The Magstim Company, Whitland, Wales, United Kingdom)ith a pediatric-size round coil (6.2 cm diameter) on the vertex. TheST seizure threshold was defined by the number of pulses

equired to elicit a seizure. Seizure threshold titration was performedy starting with a 50-Hz, 1-s stimulation (50 pulses) and deliveringtimulations approximately every 20 sec, increasing the durationy 1 sec with each stimulation until a seizure was induced.ubsequent daily dose of MST was set at 100 Hz and either 6 �eizure threshold or the maximum output capacity of the device1000 pulses at 100 Hz). The ECS was given bilaterally with auman ECT device at 2.5 � titrated seizure threshold at arequency of 50 Hz. Seizure threshold titration for ECS waserformed by increasing duration of stimulation by 160 msecith each stimulation (with current held at 800 mA and pulseidth at .5 msec) until seizure was induced. An ECS frequency of0 Hz comprises 50 pulse pairs (upward and downward going) /ec, resulting in 100 total pulses/sec. An MST frequency of 100z comprises 100 pulses/sec; thus the conditions were matched

n total number of pulses per second. The ECS electrode place-

igure 5. Task 5: Serial Probe Recognition. A list of four, five, or six stimuli isresented one by one, with no delay between stimuli. The last stimulus in aiven trial is followed by a 2-sec delay. A “probe” stimulus is then presented,nd the subject must select either a “yes” or “no” stimulus to indicatehether the probe had been in the list.

Table 1. Summary of Main Finding Comparing ECS, HD

Task Treatment C

Task 1Time to wake/begin No DifferenceTime to completion ECS � (MST � ShAccuracy No Difference

Task 2Time to completion ECS � (MST � ShAccuracy ECS � (MST � Sh

Task 3Old list accuracy ECS � (MST � ShNew list accuracy No Difference

Task 4Accuracy ECS � Sham

Task 5Accuracy ECS � (MST � Sh

HD MST, high dose magnetic seizure therapy; ECS, electro

ww.sobp.org/journal

ment was conventional bilateral (bifrontotemporal). For sham,only anesthesia was given.

Immediately after each intervention, monkeys were trans-ported to a test chamber where the neurocognitive battery wasinitiated after a response to a start stimulus on the touch screen.

Data AnalysisThe analyses used mixed effects models (MEMs) that evalu-

ated each cognitive task separately (26). Analyses were con-ducted with the PROC MIXED procedure of SAS (Cary, NorthCarolina) (27). For all tasks, fixed effects included “condition”(four levels including baseline) and “session day” (a continuousvariable). For tests administered in both morning and afternoonsessions (the T2 test and the Simchain new list test), the fixedeffect “session” (2 levels) was included. For the Simchain old listtest, in which the same list was shown on multiple consecutivedays, the fixed effect “day-of-exposure” was included (threelevels). For the spatial working memory test and SPR test, thenumber of list items (three levels) was included as a fixed effect.For Task 3, a software error midway through subject 3’s studyparticipation caused old list and new list trials to be interspersedrather than separately blocked. Because this error persistedacross multiple treatment conditions and was found to signifi-cantly reduce his Task 3 overall scores, that subject was droppedfrom that analysis. A separate analysis pooled data from thecurrent study with data from a prior, moderate-dose MST (2.5 �seizure threshold delivered at 50 Hz) study performed in thesame subjects (8). Statistical significance was judged on the basisof � � .05. The parameters were estimated with the iterativemaximum-likelihood method. For all significant findings fromcognitive test scores, effect size was calculated (Table 1).

Results

Feasibility of High-Dose MST Seizure InductionThe MST seizure was induced in all three subjects, and there

were no adverse events. The observed MST seizure thresholdswere 225 pulses for subject 1, 100 pulses for subject 2, and 300pulses for subject 3. In the MST condition, there was a meanseizure duration of 24.2 � 5.3 (SD) sec. This was significantlylonger [F (48) � .51, p � .015] than MST seizure duration in themoderate-dose study, where the mean seizure was 19.8 � 7.4(SD) sec. However, there was no significant difference betweenMST seizure duration and ECS seizure duration, which had a

, and Sham

tion Cohen’s Standard (effect size)

—Baseline Large (.7, .6, .9)

—

Baseline) Small/Medium (.2, .2, .5)Baseline) Medium (.4, .4, .4)

Baseline) Medium (.4, .4, .5)—

Small (.2)

Small (.2, .2)

MST

ondi

am) �

am �am �

am �

am)

convulsive shock; Sham, anesthesia only.

mi

T

tgso

R

cmto[Etta.M

m

C

itT.[�t

ofeEapbd

A

[

e not

T. Spellman et al. BIOL PSYCHIATRY 2008;63:1163–1170 1167

ean of 20.2 � 5.7 (SD) sec. There was no cross-study differencen the ECS group for seizure duration.

ime to Recover ConsciousnessTime from final anesthesia injection until the subject pressed

he “start” button in the cognitive testing chamber was used toauge time to recovery of consciousness (Table 2). This measurehowed no significant main effects for condition, session day, orrder.

ecall Time and Accuracy of an Over-Learned Stimulus (Task 1)Task-Completion Time. Analysis of the time required to

omplete the task (i.e., task time) once it was started, yielded aain effect of order [F (52) � 5.87, p � .0050, with the task taking

he most time in earlier treatments, indicating a decrease inverall treatment effects over time] and a main effect of condition

F (52) � 25.14, p � .0001]. Task time was significantly longer forCS than for MST, sham, or baseline [t (170) � 5.23, p � .0001;(170) � 4.40, p � .0001; t (170) � �6.83, p � .0001, respec-ively]. Baseline times were significantly shorter than both MSTnd sham times [t (170) � �2.61, p � .0100; t (170) � �3.31, p �0012, respectively], and there was no difference between theST and sham condition task times (Table 1).

Accuracy. Accuracy scores for Task 1 showed no significantain effects in the high-dose study (Table 1).

ompletion Time and Accuracy of Learning New Targets (Task 2)Task-Completion Time. There was a significant order effect

n times to task completion [F (2) � 13.86, p � .0001], with theask taking longer in the first phase than in the third (Table 3).here was also a main effect of condition [F (2) � 12.10, p �

0001], with ECS taking longer than MST, sham, or baselinet (331) � 5.15, p � .0001; t (331) � 5.35, p � .0001; t (331) �4.33, p � .0001, respectively], and no difference was found for

ask time among the MST, sham, and baseline conditions.Accuracy. Accuracy scores yielded a significant main effect

f order [F (144) � 5.43, p � .0054, with performance improvingrom the first to the last phase of the study] and a significant mainffect of condition [F (4) � 286.13, p � .0001]. Task 2 scores in theCS condition were significantly lower than in the baseline, MST,nd sham conditions [F (391) � 8.78, p � .0001; F (391) � �7.00,� .0001; F (391) � �7.09, p � .0001, respectively]. Among theaseline, MST, and sham conditions, there were no significantifferences between any pair.

ccuracy of Serial List Learning and Retention (Task 3)Old List Accuracy. Main effects of exposure were found

F (190) � 17.59, p � .0001] with scores higher on the third

Table 3. Task 2 Completion Time and Accuracy of New

Baseline HD

Completion Time (min) 1.4 (1.4) 5.2Accuracy (%) 85.9% (20.9) 80.2%

Table 2. Effects of HD MST, ECS, and Sham on Recovery

Baseline

Time to Wake (min) —Time to Task 1 Completion (min) .5 (.4)Task 1 Accuracy (%) 95.5% (8.6)

All values are reported in mean (SD) unless otherwis

Mean (SD). Abbreviations as in Table 1.

exposure than on the first. Significant differences betweenconditions [F (190) � 3.56, p � .0153] showed that ECS scoreswere significantly lower than baseline, MST, or sham scores [forbaseline, t (114) � 2.54, p � .0125; for MST, t (114) � �2.36, p �.0200; for sham, t (114) � �2.45, p � .0158], and there was nodifference between scores from any pair among baseline, MST,and sham (Figure 6).

New List Accuracy. Analyses of new list accuracy and reac-tion time scores yielded no significant main effects or interac-tions.

Spatial Working Memory (Task 4) and SPR (Task 5) AccuracySpatial Working Memory. Spatial accuracy scores yielded

significant effects for the number of list items [F (2) � 25.96,p � .0001], with higher scores found for the four-item thansix-item lists. There was a main effect of condition [F (2) �3.89, p � .0389]. The ECS scores were significantly lower thansham [t (654) � �2.46, p � .0140]. Although MST scores werehigher than ECS and sham higher than MST, these differenceswere not significant. Scores from the three treatment condi-tions did not differ significantly from baseline (Figure 7).

SPR. Accuracy scores for the SPR task yielded a main effectof condition [F (2) � 3.73, p � .0248]. The ECS scores weresignificantly lower than sham [t (567) � �2.59, p � .0099] andMST [t (567) � �2.03, p � .0429], and no difference was foundbetween MST and sham scores. Scores from the three treatmentconditions did not differ significantly from baseline.

Comparison of Moderate- and High-Dose MST Effects onCognitive Tasks

As seen in Table 4, the only significant differences betweenthe MST dosage conditions were increased start time to beginTask 1 and increased accuracy on Task 2 and Task 3 old listaccuracy in the high-dose MST relative to moderate-dose MST.

There was a main effect of study on start time to begin Task1, with wake/start time significantly longer in the high-dose thanin the moderate-dose study [F (1) � 9.33, p � .0026], and acondition � study interaction [F (6) � 34.35, p � .0001]. Therewas no study effect for the ECS group (which was provided at theidentical dosage between the two studies), but both the MST andsham groups took significantly longer time to recover in thehigh-dose study than in the moderate-dose study [t (260) ��3.56, p � .0004; t (260) � �2.31, p � .0217, respectively].

For Task 2 accuracy, there was a significant main effect ofstudy, with accuracy scores in the current study higher than thosein the moderate-dose study [F (1) � 25.05, p � .0001], and acondition � study interaction [F (8) � 180.45, p � .0001]. This

t Learning Between Treatment Conditions

ECS Sham p

15.2 (28.3) 4.3 (3.2) �.0001) 56.4% (28.5) 80.3% (23.9) �.0001

rientation and Task 1 Performance

MST ECS Sham p

4 (11) 12 (9) 12 (10) ns3 (2) 8 (3) 4 (2) .005

(17.9) 72.9% (19.6) 73.0% (17.0) ns

ed. Abbreviations as in Table 1.

Targe

-MST

(7.9)(22.7

of O

HD

1

76.0%

www.sobp.org/journal

ioc.m

sM1m

D

h(sismhe

pafcrAdlda

dMdrnwehr

1168 BIOL PSYCHIATRY 2008;63:1163–1170 T. Spellman et al.

w

ncrease in scores between studies was not found in the baseliner ECS conditions, but it was significant in the MST and shamonditions [F (600) � �3.86, p � .0001; F (600) � �4.50, p �0001, respectively]. For time to complete Task 2, there was noain effect of study and no study � condition interaction.On Task 3 old list recall, there was a significant condition �

tudy interaction [F (3) � 3.19, p � .0250]. Specifically, high-doseST showed higher accuracy than moderate-dose MST [t (293) �.98, p � .0488]. There were no differences between high- andoderate-dose MST on Task 3 new list learning.

iscussion

We have shown for the first time that chronic treatment withigh-dose MST resulted in less cognitive impairment than ECSsee Table 1 for summary) and that high-dose MST did notignificantly differ from the effects of anesthesia alone. Moreover,ncreasing MST dosage from 2.5 � seizure threshold to 6 �eizure threshold did not impair cognitive performance on mosteasures. These results support the feasibility and safety ofigh-dose MST for subsequent work in humans to assess itsfficacy in the treatment of depression.

Replicating our prior report, we found that ECS impairederformance on anterograde amnesia (Task 2) and retrogrademnesia (Task 3 old list), whereas high-dose MST did not differrom sham on these measures (8,10). This result supports theonclusion that high-dose MST given at 6 � seizure thresholdetains cognitive advantages relative to ECS on these measures.lthough we failed to replicate our previously reported finding of aifference between ECS and moderate-dose MST on the newist-learning component of Task 3, neither ECS nor high-dose MSTiffered from sham on this task in the current study, likely owing tohigh degree of inter-day variance (mean � 42% � 27%).As an additional test of the cognitive impact of increasing MST

ose, we compared the current results with 100-Hz high-doseST with results seen in our prior study with 50-Hz moderate-ose MST (8). On the measure of reorientation time (the timeequired to complete Task 1 upon waking), high-dose MST didot differ from moderate-dose MST, suggesting that MST doseithin the range tested is unrelated to reorientation time. Furthervidence for rapid reorientation after MST has been seen inuman studies, which found ECT to result in significantly longer

eorientation time than acute (9) and chronic (12,13) low-dose

ww.sobp.org/journal

MST. This difference might be clinically meaningful, becauseprolonged time to orientation is correlated with retrogradeamnesia (11). More recently, a report on the first 11 depressedpatients to receive 100-Hz MST likewise found high-dose MST toresult in faster orientation recovery than ECT (28).

Interestingly, we found improved accuracy on Task 2 andTask 3 old list recall in the high-dose compared with themoderate-dose study for the MST and sham groups but not forECS condition. Task 3 old list accuracy also improved signifi-cantly in the second study for the MST group alone. A possibleexplanation is a practice effect that allowed subjects to adapt totesting after recovery from anesthesia. The two subjects in themoderate-dose study were experimentally naïve, whereas thethree subjects in the current high-dose study had collectivelyreceived numerous treatment days involving methohexital anes-thesia and subsequent cognitive testing. The fact that the ECScondition failed to show this practice effect provides furtherevidence for ECS-related impairment not seen in the MST con-ditions.

The apparent lack of MST dose dependence—within therange tested here—on learning and memory functions assessed

Figure 6. Task 3: Sequence memory. For previously learned(old) lists, electroconvulsive shock (ECS) resulted in signifi-cantly poorer performance than in baseline, magnetic seizuretherapy (MST), or sham (*p � .02). For new lists, ECS and MSTdid not differ from sham.

Figure 7. In Task 4, the spatial working memory task, electroconvulsiveshock (ECS) resulted in significantly lower scores than sham (*p � .0140). InTask 5, the serial probe recognition task, ECS resulted in lower scores than

magnetic seizure therapy (MST) or sham (*p � .05).

bcm1Bwabosdfsoilcar

twssmbpidwimmmfwhpMofpasdn

T

T

T

T

T

ich bor

T. Spellman et al. BIOL PSYCHIATRY 2008;63:1163–1170 1169

y Tasks 1, 2, and 3 warrants further discussion. Both MSTonditions were given at the same intensity (e.g., strength of theagnetic field), and they differed only in frequency (50 versus

00 Hz) and the duration of each stimulation train (in seconds).ecause the induced electric field generated by individual pulsesithin the two conditions was identical, the depth of penetrationnd regions of the brain directly stimulated by each pulse shoulde identical. We have previously reported the depth of penetrationf MST to be superficial compared with ECS (6). Thus, repeatedlytimulating superficial cortex at higher frequencies and longer trainurations would not be expected to adversely affect cognitiveunctions subserved by brain regions remote from the site oftimulation, except by transsynaptic action. The learning and mem-ry functions involved in Tasks 1, 2, and 3 might be expected tonvolve hippocampal regions, which would not be directly stimu-ated even in the high-dose MST condition. However, our dosageomparison was retrospective and confounded by subject age; thusdefinitive test of the dose-response relationships will require

andom assignment to different MST dose levels.In this study we introduce an expanded version of the CUPCP

hat includes two new measures of working memory: 1) spatialorking memory (Task 4), and 2) SPR (Task 5). The ECS

ignificantly impaired working memory on both tasks relative toham, whereas high-dose MST did not. The spatial workingemory task was based upon the radial arm maze, a task that haseen reported to result in increased activation of the hippocam-us (29). Because previous research has shown that MST results

n less marked physiological and anatomical changes in theentate gyrus than ECS (6,30), we predicted high-dose MSTould have less of an effect on this task than ECS. Further work

s needed to establish the extent to which this task, as imple-ented here, provides a measure of hippocampal functioning inonkeys. The SPR task was modeled after the Sternberg delayedatch-to-sample task, which has been reported to activate

rontal, parietal, occipital, and other cortical regions in humans asell as deeper structures such as cingulate cortex, insula, andypothalamus and to be associated with sustained firing inarahippocampal neurons (31–33). We expected that high-doseST might impair performance on this task by virtue of its actionn superficial cortex. However, high-dose MST did not impairunction on this task relative to sham and fared better than ECS,erhaps again owing to their differential impact on areas that arective during the task but that fall outside the area most focallytimulated by MST. However, definitive interpretation of theseifferences awaits further study of the neurobiological underpin-

able 4. Comparison of High- and Moderate-Dose MST on Cognitive Effec

ask HD MST (3 subjects) Mod-Dose M

ask 1Time to wake/begin (min)a 14 (11)Time to completion (min) 3 (2)Accuracy (%) 76.0% (18.0) 85.4

ask 2Time to completion (min) 5.2 (7.9) 8Accuracy (%) 80.2% (22.7) 54.2

ask 3Old list accuracy (%) 63.8% (20.0) 43.6New list accuracy (%) 39.1% (25.9) 29.5

All values are expressed as mean (SD) unless otherwise noted. HD MST,aThe time to wake/begin showed a study � condition interaction in wh

elative to the old study.

ings of these tasks in primates (e.g., via functional imaging).

Limitations of this study include the small sample size, thelimited number of cognitive domains assessed, the risk ofcarry-over effects, practice effects, the fact that all three subjectswere male, and the aforementioned confound between age andMST dosage in the retrospective dosage comparisons. A largersample size would be needed before these results could begeneralized. Additionally, because this is the first report of thetwo new working memory tasks added to the CUPCP, furtherwork will be needed to characterize and validate their psychometricproperties. A further limitation is the use of bilateral ECS rather thanunilateral, which might be expected to have fewer cognitive sideeffects. However, bilateral ECS was selected to match the MSTplacement, which was also bilateral. It should also be noted that thedosage relative to seizure threshold was imbalanced between thetwo conditions. Specifically, MST was given at 6 � seizure thresh-old, whereas ECS was given at 2.5 � seizure threshold. This dosageimbalance could be expected to bias the study in favor of seeing anadvantage of ECS. Our results were the opposite, further strength-ening the support for the relative safety of MST.

For animal models of cognitive function to meaningfullyinform the development of safer neurostimulation interventionsfor patient populations, it will be important to develop systematictranslational techniques that allow for cross-species comparisonof neurocognitive function. A potentially useful approach to suchtranslational research would be the development of parallelanimal and human models of specific neurocognitive domains(34). Such an approach allows one to leverage invasive measuresof anatomy and physiology available in the animal to elucidatemechanisms of action and the specific neurocircuitry affected byneurostimulation techniques (35). Moreover, the parallel exami-nation of human and nonhuman responses to neurostimulationcan shed light on similar and divergent cognitive processesacross species. As we learn more about the effects of varioustypes of neurostimulation modalities, controlling for variablessuch as stimulation focality; current density; and depth, strength,and duration of seizure propagation, the neurocognitive effectsof these types of stimulation can provide insight into the roles ofanatomy and cross-species divergence in cognition.

This study was supported by National Institute of MentalHealth (NIMH) R01 MH60884. Support for the development ofmagnetic seizure therapy has also come from the Stanley MedicalResearch Foundation, American Federation for Aging Research /Beeson Scholars Program, and the National Alliance for Re-

subjects) Statistic p Effect Size (Cohen’s d)

t(260) � �3.56 .0004 Medium (.45)ns —

.3) ns —

) ns —.1) F(600) � �3.86 .0001 Medium (.45)

.7) t(293) � 1.98 .0488 Medium (.41)

.8) ns —

dose magnetic seizure therapy; Mod-Dose, moderate dose.th the MST and Sham conditions showed lengthier times in the new study

ts

ST (2

6 (2)3 (2)

% (15

.2 (8.8% (29

% (24% (19

high-

search on Schizophrenia and Depression (NARSAD).

www.sobp.org/journal

OC

rCNvohstMstoZGoL

1

1

1

1

1170 BIOL PSYCHIATRY 2008;63:1163–1170 T. Spellman et al.

w

The authors thank Dr. Tammy Moscrip, Dr. Mohammedsman, Niko Reyes, Nick DeWind, and the members of theolumbia Primate Cognition Laboratory.

For work unrelated to the present study, Dr. Lisanby haseceived research support from Magstim Company, Neuronetics,yberonics, NIH, American Federation for Aging Research,ARSAD, Stanley Medical Research Foundation, Defense Ad-anced Research Projects Agency, and the New York State Officef Science, Technology, and Academic Research. Dr. McClintockas received funding from NIH. Dr. Husain has received re-earch support from the NIMH, Stanley Medical Research Insti-ute, Cyberonics, Pfizer (in process), Neuronetics, Magstim, andedtronics (potential research sponsor); he has served on Advi-

ory Boards for AstraZeneka, VersusMed, Avinar, Boston Scien-ific, MEASURE, Bristol-Meyer-Squibb; and Clinical Advisors andn speakers bureaus for Cyberonics, Avinar, Cerebrio, Astra-eneka, Bristol-Meyers-Squibb, Optima/Forrest Pharmaceuticals,laxo-Smith-Kline, Forrest Pharmaceuticals, and Janssen. Tim-thy Spellman, Herb Terrace, Shawn McClintock, and Bruceuber have no conflicts or potential conflicts to report.

1. American Psychiatric Association. Task Force on Electroconvulsive Ther-apy (2001): The Practice of Electroconvulsive Therapy: Recommendationsfor Treatment, Training, and Privileging, 2nd ed. Washington, DC: Ameri-can Psychiatric Association.

2. Sackeim HA, Prudic J, Fuller R, Keilp J, Lavori PW, Olfson M (2007): Thecognitive effects of electroconvulsive therapy in community settings.Neuropsychopharmacology 32:244 –254.

3. Lisanby S, Maddox J, Prudic J, Devanand D, Sackeim H (2000): The effectsof electroconvulsive therapy on memory of autobiographical and pub-lic events. Arch Gen Psychiatry 57:581–590.

4. Lisanby SH, Schlaepfer TE, Fisch HU, Sackeim HA (2001): Magnetic sei-zure therapy of major depression. Arch Gen Psychiatry 58:303–305.

5. Lisanby SH, Luber B, Finck AD, Schroeder C, Sackeim HA (2001): Deliber-ate seizure induction with repetitive transcranial magnetic stimulationin nonhuman primates. Arch Gen Psychiatry 58:199 –200.

6. Lisanby SH, Moscrip TD, Morales O, Luber B, Schroeder C, Sackeim HA(2003): Neurophysiological characterization of magnetic seizure ther-apy (MST) in nonhuman primates. Suppl Clin Neurophysiol 56:81– 89.

7. Lisanby SH, Peterchev AV (2007): Magnetic seizure therapy for the treat-ment of depression. In: Marcolin MA, Padberg F, editors. TranscranialBrain Stimulation for Treatment of Psychiatric Disorders. Basel, Switzer-land: Karger Publishers, 155–171.

8. Moscrip TD, Terrace HS, Sackeim HA, Lisanby SH (2006): Randomized,controlled trial of the cognitive side-effects of magnetic seizure therapy(MST) and electroconvulsive shock (ECS). Int J Neuropsychopharmacol8:1–11.

9. Lisanby SH, Luber B, Schlaepfer TE, Sackeim HA (2003): Safety and feasi-bility of magnetic seizure therapy (MST) in major depression: Random-ized within-subject comparison with electroconvulsive therapy. Neuro-psychopharmacology 28:1852–1865.

0. Moscrip T, Terrace HS, Sackeim HA, Lisanby SH (2004): A primate modelof anterograde and retrograde amnesia produced by convulsive treat-ment. J ECT 20:26 –36.

1. Sobin C, Sackeim HA, Prudic J, Devanand DP, Moody BJ, McElhiney MC(1995): Predictors of retrograde amnesia following ECT. Am J Psychiatry152:995–1001.

2. Lisanby SH, Husain M, Morales O, Thornton L, White PF, Payne N, et al.(2003): Controlled clinical trial of the antidepressant efficacy of mag-netic seizure therapy in the treatment of major depression. ACNP An-nual Meeting Abstracts.

3. White PF, Amos Q, Zhang Y, Stool L, Husain MM, Thornton L, et al. (2006):

Anesthetic considerations for magnetic seizure therapy: A novel ther-apy for severe depression. Anesth Analg 103:76 – 80.

ww.sobp.org/journal

14. McClintock SM, Husain MM, Cullum CM, Luber B, Lisanby SH (2007):Neurocognitive associated effects of magnetic seizure therapy. ClinNeuropsychologist 21:399.

15. Sackeim HA, Prudic J, Devanand D, Kiersky JE, et al. (1993): Effects ofstimulus intensity and electrode placement on the efficacy and cogni-tive effects of electroconvulsive therapy. New Engl J Med 328:839 – 846.

16. Ng C, Schweitzer I, Alexopoulos P, Celi E, Wong L, Tuckwell V, et al.(2000): Efficacy and cognitive effects of right unilateral electroconvul-sive therapy. J ECT 16:370 –379.

17. Ward WK, Lush P, Kelly M, Frost AD (2006): A naturalistic comparison oftwo right unilateral electroconvulsive therapy dosing protocols: 2–3Xseizure threshold versus fixed high-dose. Psychiatry Clin Neurosci 60:429 – 433.

18. Sackeim HA, Prudic J, Devanand D, Nobler MS, Lisanby SH, Peyser S, et al.(2000): A prospective, randomized, double-blind comparison of bilat-eral and right unilateral electroconvulsive therapy at different stimulusintensities. Arch Gen Psychiatry 57:425– 434.

19. McCall WV, Dunn A, Rosenquist PB, Hughes D (2002): Markedly suprath-reshold right unilateral ECT versus minimally suprathreshold bilateralECT: Antidepressant and memory effects [see comment]. J ECT 18:126 –129.

20. McCall WV, Reboussin DM, Weiner RD, Sackeim HA (2000): Titratedmoderately suprathreshold vs fixed high-dose right unilateral electro-convulsive therapy: Acute antidepressant and cognitive effects. [seecomment]. Arch Gen Psychiatry 57:438 – 444.

21. Peterchev A, Kirov G, Ebmeier K, Scott A, Husain M, Lisanby S (2007):Frontiers in TMS technology development: Controllable pulse shapeTMS (cTMS) and magnetic seizure therapy (MST) at 100 Hz. Biol Psychia-try 61:107S.

22. Gavan JA, Swindler DR (1966): Growth rates and phylogeny in primates.Am J Phys Anthropol 24:181–190.

23. Tigges J, Gordon T, McClure H, Hall E, Peters A (1988): Survival rate andlife span of rhesus monkeys at the Yerkes Regional Primate ResearchCenter. Am J Primatol 15:263–273.

24. Hackers S, Zalesak M (2004): Hippocampal activation during transitiveinference in humans. Hippocampus 14:53–162.

25. Sternberg S (1966): High-speed scanning in human memory. Science153:652– 654.

26. Diggle PJ HP, Liang KY, Zeger SL (2002): Analysis of Longitudinal Data,2nd ed. Oxford: Oxford University Press.

27. Littell R MG, Stroup W, Wolfinger R (1996): SAS System for Mixed Models.Cary, North Carolina: SAS Institute.

28. Kirov G, Ebmeier KP, Scott A, Atkins M, Khalid N, Carrick L, et al. (in press):Quick recovery of orientation after 100 Hz magnetic seizure therapy(MST) for major depressive disorder. Br J Psychiatry.

29. Astur RS, Germain SA, Baker EK, Calhoun V, Pearlson GD, Constable T(2005): fMRI hippocampal activity during a virtual radial arm maze.Applied Psychophysiol Biofeedback 30:307–317.

30. Scalia J, Lisanby S, Underwood M, Sackeim H, Dwork A, Morales O, et al.(2004): The spatial distribution of mossy fiber sprouting in a non-humanprimate model for electroconvulsive therapy and magnetic seizuretherapy. Biol Psychiatry 55:207S.

31. Walter H, Wolf RC, Spitzer M, Vasic N (2007): Increased left prefrontalactivation in patients with unipolar depression: An event-related, para-metric, performance-controlled fMRI study. J Affect Disord 101:175–185.

32. Schon K, Atri A, Hasselmo ME, Tricarico MD, LoPresti ML, Stern CE (2005):Scopolamine reduces persistent activity related to long-term encodingin the parahippocampal gyrus during delayed matching in humans.J Neurosci 25:9112–9123.

33. Jansma JM, Ramsey NF, de Zwart JA, van Gelderen P, Duyn JH (2007):fMRI study of effort and information processing in a working memorytask. Hum Brain Mapp 28:431– 440.

34. Rosen JB (2006): Translational research: Parallels of human and animalresearch in biological psychology. Biol Psychology 73:1–2.

35. Strauman TJ, Merrill KA (2004): The basic science/clinical science inter-

face and treatment development. Clin Psychol Sci Pract 11:263–266.