Education Research: Cognitive performance is preserved in sleep

Education Research:Cognitive performance is preserved insleep-deprived neurology residents

M. Reimann, PhDR. Manz, MDS. PrieurH. Reichmann, ProfT. Ziemssen, MD

ABSTRACT

Objective: To test the hypotheses that sleep deprivation in neurology residents is associated withperformance deficits and that vigilance and cognitive performance is more compromised afterovernight on-call duty compared to night shift.

Methods: Thirty-eight neurology residents of a university teaching hospital participated in a prospec-tive single-blind comparison study. Residents were recruited according to their working schedule anddivided into 3 groups: 24 hours overnight on-call duty, night shift, and regular day shift (controls). Allparticipants underwent serial measurements of sleepiness and cognitive performance in the morningdirectly after or before their shift. Pupillary sleepiness test and Paced Auditory Serial Addition Testwere applied. Perceived sleepiness was assessed by a questionnaire.

Results: Sleepiness was increased in residents after night shift and overnight call compared tocontrols while the type of night duty was not associated with the extent of sleepiness. Sleep-deprived residents did not show any performance deficits on the Paced Auditory Serial AdditionTest. Cognitive performance was not associated with sleepiness measures.

Conclusions: Night shift and overnight call duty have a similar impact on alertness in neurologyresidents. Sleep-deprived neurology residents may be able to overcome sleep loss–related per-formance difficulties for short periods. Neurology® 2009;73:e99 –e103

GLOSSARYPASAT � Paced Auditory Serial Addition Test; PST � Pupillography Sleepiness Test; PUI � pupillary unrest index; SSS �self-stated sleepiness.

Despite recent changes in working schedule regulation for clinicians, long working hours remain acommon feature in health care.1-3 Associated sleep deprivation and fatigue present not only a seriousconcern for patient safety, but also places the health of health care professionals at risk. Seriousproblems resulting from sleep loss range from performance deficits and erroneous decision makingto increased risk for motor vehicle accidents.2,4

The majority of sleepiness studies have been performed in certain medical specialties with areputation for demanding schedules, such as surgery or intensive care.2 Neurologists are frequentlyunderrated in terms of intensity and heaviness of their working schedules due to their mistakenlyclose relationship to psychiatry. Call rotation on the neurology ward and night shifts at the neurol-ogy intensive care unit are as common as in internal medicine or surgery. Actually, neurologists arefrequently challenged by the high prevalence of life-threatening strokes in the Western society,where any delay in action or poor performance may be fatal for the patient.

This prompted us to investigate sleepiness and cognitive performance in sleep-deprived and alertneurologists working at a large university neurology clinic. We hypothesized that 1) sleep depriva-tion in neurologists is associated with performance deficits as previously demonstrated for other

Address correspondence andreprint requests to Dr. ManjaReimann, ANF Laboratory,Department of Neurology,University Clinic Carl GustavCarus, Dresden University ofTechnology, Fetscherstraße 74,D-01307 Dresden, [email protected]

From the Autonomic and Neuroendocrinological Laboratory, Department of Neurology (M.R., R.M., S.P., H.R., T.Z.), and Research Group Neuro-Metabolism, Department of Neurology and Internal Medicine III (M.R., T.Z.), University Hospital Carl Gustav Carus, Dresden University ofTechnology, Germany.

Supported by the University Hospital Carl Gustav Carus, Dresden, Germany.

Disclosure: Author disclosures are provided at the end of the article.

RESIDENT& FELLOWSECTION

Section EditorMitchell S.V. Elkind,MD, MS

Copyright © 2009 by AAN Enterprises, Inc. e99

medical specialties and that 2) vigilance as wellas cognitive performance is more compromisedafter 24 hours overnight on-call duty comparedto night shift.

METHODS This pilot study included 38 neurology residents(19 women and 19 men, mean age 30 � 2 years) recruited from theDepartment of Neurology at the University Hospital Carl GustavCarus, Dresden, Germany. We screened residents for exclusion cri-teria such as current use of medication known to affect the sleep/wake cycle or daytime alertness, current psychiatric illness, and sleepdisorder diagnosis. We then stratified eligible residents according totheir working schedule into 3 groups: group 1, 24 hours overnighton-call duty; group 2, night shift; and group 0, regular day shift(control).

Definitions of groups. Residents in group 1 (24 hours over-night on-call duty) performed their regular shift from 8 AM–4PM followed by overnight call until 8 AM the next day. Duringtheir overnight duty, they were allowed to sleep but there was noscheduled coverage. Residents have usually 3 to 4 overnight callsper month. Residents in group 2 (night shift) worked at theintensive care unit for 7 consecutive days daily from 8 PM to 8 AM

the following morning. Afterwards they had 1 week off. Sleepingwas not permitted during their shifts. From experience, residentshave on average 2 admissions and 30 internal and 3 outpatientconsultations per night. Night shift rotation was every 5 to 6weeks over a period of 1 year. Residents in group 0 (day shift)regularly worked from 7:30 AM to 3:30 PM. However, residentson day shift frequently work overtime.

Standard protocol approvals, registrations, and patient con-sents. The Ethics Committee on human experimentation of theDresden University of Technology approved the study and the in-vestigation conformed with the principles outlined in the Declara-tion of Helsinki. We informed all participating residents about theobjectives and procedures of the study and obtained written in-formed consent prior to their inclusion. The serial data collectionwas performed at the Autonomic and Neuroendocrinological Labo-ratory of the University Clinic. We measured all residents before 9AM directly after their (night) shift rotation or just before day shiftcommenced (controls). The assistant performing the measurementswas blinded. We measured objective sleepiness by PupillographySleepiness Test (PST) and cognitive performance by Paced Audi-tory Serial Addition Test (PASAT). We instructed the residents toabstain from drinking alcoholic beverages, smoking, and drinkingcoffee for at least 4 hours before the measurements. The residentsrated their sleepiness on a 5-point Likert scale based on the state-ment “Currently I feel.” We also recorded the number of hours sleptin the previous 24 hours and assessed the perceived recovery effect

due to sleep on a 5-point Likert scale. All measurements were seriallyrepeated up to a maximum of 13 times. Measurement frequency,however, varied between individuals according to the rotationschedules (4.8 � 3.3).

Pupillary Sleepiness Test. We performed the PST (AMTech,Dossenheim, Germany) in a quiet and darkened room after an ini-tial dark-adapting phase of 15 minutes. During PST, residents woregoggles equipped with infrared light transmitting filter glasses im-pervious to visible light. They were seated on a comfortable chairand head position was adjusted by a chin rest fixed on a table. Aninfrared video camera was fixed at a distance of 70 cm from theexamination subject. We instructed the clinicians to maintain fixa-tion on a set of infrared light-emitting diodes. We then recordedspontaneous pupillary oscillations over a period of 11 minutes byinfrared video pupillography and evaluated the recording by 25-Hzreal-time analysis as published elsewhere.5 Pupillary unrest index(PUI) is a measure of pupillomotor hippus in darkness and calcu-lated as an integrated sum of slow movements of the pupillary mar-gin during the measurement period.6 This value is usually low inalertness and increases with progressive sleepiness. We also calcu-lated the mean pupil diameter over the entire recording period of 11minutes. During sleepiness the initial diameter is reduced and themean pupil size falls below the initial diameter toward the end of themeasurement.

Paced Auditory Serial Addition Test. The validated andcomputer-aided PASAT allows for measuring the capacity andvelocity of information processing within the auditory-verbal do-main (cognitive performance).7,8 The test system entails the sub-ject to continuously add the last 2 numbers of consecutive seriesand to announce the sum aloud. Numbers from 1 to 9 are an-nounced acoustically in random order by a PC with the screenremaining dark. To avoid practice effects, clinicians were trainedon the PASAT at least 3 times before commencing the study. Weapplied the 60-item short version of the test (maximal score of60). Lower scores (small numbers of correct answers) indicateworse cognitive performance.

Statistical analysis. We used the SPSS software package ver-sion 16.0 for Windows (SPSS Inc., Chicago, IL) for all statisticalevaluation. Data are presented as median and 25th–75th percen-tile unless otherwise stated. Owing to the small sample size weassumed non-Gaussian distribution, and hence applied nonpara-metric tests with Bonferroni correction for comparing groups.Spearman correlation coefficients were calculated. A two-tailedp � 0.05 was regarded as the level of significance.

RESULTS Before we started the comparison amongthe 3 groups, we assessed the strength of associationbetween the first measurement and the mean of serial

Table Sleepiness and cognitive performance of neurology residents by type of night duty

Parameters Overnight call (n � 17) Night shift (n � 6) Control (n � 15) �2 (p*)

Pupil diameter (mm) 7.51 (6.25–7.75) 7.23 (5.18–7.54) 7.21 (6.61–8.00) 0.59 (0.744)

Pupillary unrest index (mm/min) 7.00 (4.96–9.44)a 10.34 (7.78–15.07)a 4.72 (3.86–5.09)b 12.88 (0.002)

PASAT score† 56 (49–58) 51 (42–58) 54 (48–56) 0.67 (0.715)

Self-stated sleepiness‡ 2.3 (1.7–2.7)a 2.6 (2.2–3.1)a 1.0 (0–1.0)b 20.24 (�0.001)

Data are median (interquartile range). Unequal superscript letters indicate significant differences (Mann-Whitney U test).*Kruskal-Wallis test.†Paced Auditory Serial Addition Test (number of correct answers/60).‡Categorical value.

e100 Neurology 73 November 24, 2009

measurements. We revealed significant correlationswith r values ranging from 0.763 to 0.904 (p �0.001). Based on these results, we continued ouranalysis using the mean values.

PUI and self-stated sleepiness (SSS) were signifi-cantly affected by type of night duty while pupil diame-ter and the PASAT score remained unaffected (table). Itappeared that PUI and SSS were significantly higherafter the night shift and the 24 hours overnight on-callduty compared to a normal night at home. Residentsafter night shift and 24 hours overnight on-call duty didnot differ with respect to sleepiness measures.

Neurology residents on 24 hours overnight on-call duty had slept on average 4.3 (2.8–4.6) hours(midshift nap) in the last 24 hours, which was signif-icantly less compared to their colleagues on nightshift (5.9 [4.9–7.0] hours, p � 0.006) or on day shift(controls) (6.5 [6.0–7.0] hours, p � 0.001). Thelongest sleeping phase during 24 hours overnight on-call duty was 3.0 (2.0–3.8) hours. Residents had on

average a mean (minimum–maximum) of 2 (1–3)admissions, 2 (0–7) consultations, and 3 (1–5) tele-phone inquiries during overnight on-call duty.

Figure 1 illustrates the proportion of respectiveresponses to the statement “Currently I feel . . .” Fig-ure 2 depicts the self-stated recovery effect due tosleep in the 24 hours preceding the examination.

Correlation analyses did not reveal any associationbetween the PUI and the PASAT score. However, thePUI increased (r � 0.507, p � 0.001) and the PASATscore decreased (r � �0.335, p � 0.04) with increasedSSS in the total sample. The perceived level of sleepinessdecreased as the number of sleeping hours in the past 24hours increased (r � �0.527, p � 0.001). The aboveassociations could not be confirmed in subgroups (p �

0.05).

DISCUSSION Rotating shift work in clinics to pro-vide 24-hour patient care has come increasingly underscrutiny due to negative effects associated with sleeploss, fatigue, and circadian disruption.9,10 Althoughnight shift and 24 hours overnight on-call duty consid-erably differ in terms of number of working hours, per-mission for midshift naps, and rotation frequency, theireffect on sleepiness and cognitive performance has neverbeen distinguished. We hypothesized that residents on24 hours call rotation would be more affected by sleeploss due to a longer and more irregular working sched-ule. We investigated this hypothesis in neurology resi-dents of a large university clinic. This specialty group isoften underrated in terms of heaviness and intensity oflabor and therefore has never been investigated in sleep-iness studies. Importantly, previous sleepiness studies inselected medical specialties explicitly emphasized thatresults must not be extrapolated to other medical spe-cialty groups.3,4,11 Although we could not verify theabove hypothesis, our results clearly demonstrate thatsleepiness is a common problem among neurology resi-dents undergoing night shift and 24 hours overnighton-call duty.

We additionally demonstrated that vigilance mea-sured by PST is in good agreement with SSS. Thisfinding corresponds with previous studies suggestingthe PST as a valid and objective tool to detect sleepi-ness in healthy subjects.6

The lack of significant performance decrements insleep-deprived neurology residents vs controls is intrigu-ing since an association between performance and acutesleep deprivation was found in previous studies.2,4,12

Differences in study design, medical specialty, andmethods for vigilance testing limit comparisons acrossstudies. However, one study also failed to show a signif-icant performance decrement on the complex PASATtest in sleep-deprived normal subjects.13 The investiga-tors concluded that university-based research may func-

Figure 1 Self-stated sleepiness of neurology residents by type of night duty

Figure 2 Self-stated recovery effect due to sleep in the past 24 hours

Neurology 73 November 24, 2009 e101

tion as a motivational incentive, which tends to offsetsleepiness effects on performance.13

Alternatively, the applied method for performancetesting may have been suboptimal for our populationsince its clinical utility is mainly proven in neuropsycho-logical syndromes. However, more recent studies alsoemployed the PASAT in healthy adults with satisfactoryresults.7,14 Originally assumed to measure rate of infor-mation processing, the PASAT is now recognized astapping into different types of cognitive processes.15

This multifactorial nature may complicate the interpre-tation of test result from sleep-deprived subjects espe-cially under the assumption that sleep loss affectsdifferent cognitive pathways differentially. The consis-tently high PASAT score across all study groups sug-gests that sleep-deprived neurology residents are able toovercome performance difficulties for a limited time pe-riod. This assumption seems plausible since stressfulworking situations are routine in the clinic and thusmay facilitate partial conditioning. Although there is ev-idence that sleep-deprived resident doctors are prone tocommitting medical errors,4,9 tasks of short durationmay be less likely to detect performance deficits inchronically sleep deprived individuals.2

Importantly, we adequately controlled for prac-tice effects8 first by using a control group and sec-ondly by training the clinicians on the PASAT priorto the study. Consequently, we obtained a very goodagreement between the first measurement and themean of serial measurements. In addition, majorconfounders of the PASAT such as age, gender, edu-cation, and ethnicity14,16 were accounted for by creat-ing a homogenous study sample.

Nevertheless, it must be considered that our resultsare compromised by a questionable rested controlgroup. Correspondingly, the average amount of sleep inour controls was equivalent to the accepted core sleeprequirement of 6.5 hours (many were also below).17 Inaddition, nearly two thirds of the controls felt fairly topoorly rested, which indicates chronic partial sleep de-privation (figure 3). This finding is relevant as chronicpartial sleep deprivation appears to be cumulative withrespect to performance decrements.18

Neurology residents on night shift and overnightcall are affected to a similar extent by sleepiness. In-creased sleepiness, however, did not affect performanceon the complex PASAT test. It seems that sleep-deprived neurology residents may be able to overcomesleep loss–related performance difficulties for short peri-ods. This, however, may not necessarily apply for moredemanding procedures outside the laboratory.

AUTHOR CONTRIBUTIONSStatistical analysis was conducted by Dr. Manja Reimann.

ACKNOWLEDGMENTThe authors thank all neurology residents for their participation in this study.

DISCLOSUREDr. Reimann, Dr. Manz, and S. Prieur report no disclosures. Dr. Reich-

mann serves on scientific advisory boards, receives speaker honoraria,

and/or receives funding for travel from Cephalon, Inc., Novartis, Teva

Pharmaceutical Industries Ltd., Lundbeck Inc., GlaxoSmithKline, Boehr-

inger Ingelheim, Bayer Schering Pharma., UCB/Schwarz Pharma, Desitin

Pharmaceuticals, GmbH, Pfizer Inc., and Solvay Pharmaceuticals, Inc.

Dr. Ziemssen has received speaker honoraria from Biogen Idec, Sanofi-

Aventis, Merck Serono, Novartis, Teva Pharmaceutical Industries Ltd.,

and Bayer Schering Pharma; serves as a consultant for Teva Pharmaceuti-

cal Industries Ltd., Novartis, and Bayer Schering Pharma; and receives

research support from the Roland Ernst Foundation.

REFERENCES1. Reddy R, Guntupalli K, Alapat P, Surani S, Casturi L,

Subramanian S. Sleepiness in medical ICU residents.Chest 2009;135:81–85.

2. Veasey S, Rosen R, Barzansky B, Rosen I, Owens J. Sleeploss and fatigue in residency training: a reappraisal. JAMA2002;288:1116–1124.

3. Surani S, Subramanian S, Aguillar R, Ahmed M, Varon J.Sleepiness in medical residents: impact of mandated reduc-tion in work hours. Sleep Med 2007;8:90–93.

4. Arnedt JT, Owens J, Crouch M, Stahl J, Carskadon MA.Neurobehavioral performance of residents after heavy nightcall vs after alcohol ingestion. JAMA 2005;294:1025–1033.

5. Ludtke H, Wilhelm B, Adler M, Schaeffel F, Wilhelm H.Mathematical procedures in data recording and processing ofpupillary fatigue waves. Vision Res 1998;38:2889–2896.

6. Wilhelm B, Wilhelm H, Ludtke H, Streicher P, Adler M.Pupillographic assessment of sleepiness in sleep-deprivedhealthy subjects. Sleep 1998;21:258–265.

7. Diehr MC, Cherner M, Wolfson TJ, Miller SW, Grant I,Heaton RK. The 50 and 100-item short forms of thePaced Auditory Serial Addition Task (PASAT): demo-graphically corrected norms and comparisons with the fullPASAT in normal and clinical samples. J Clin Exp Neuro-psychol 2003;25:571–585.

8. Tombaugh TN. A comprehensive review of the Paced Au-ditory Serial Addition Test (PASAT). Arch Clin Neuro-psychol 2006;21:53–76.

9. Barger LK, Cade BE, Ayas NT, et al. Extended work shiftsand the risk of motor vehicle crashes among interns.N Engl J Med 2005;352:125–134.

10. Gold DR, Rogacz S, Bock N, et al. Rotating shift work,sleep, and accidents related to sleepiness in hospital nurses.Am J Public Health 1992;82:1011–1014.

11. Saxena AD, George CF. Sleep and motor performance in on-call internal medicine residents. Sleep 2005;28:1386–1391.

12. Grantcharov TP, Bardram L, Funch-Jensen P, RosenbergJ. Laparoscopic performance after one night on call in asurgical department: prospective study. BMJ 2001;323:1222–1223.

13. Hood B, Bruck D. A comparison of sleep deprivation andnarcolepsy in terms of complex cognitive performance andsubjective sleepiness. Sleep Med 2002;3:259–266.

14. Diehr MC, Heaton RK, Miller W, Grant I. The PacedAuditory Serial Addition Task (PASAT): norms for age,education, and ethnicity. Assessment 1998;5:375–387.

15. Madigan NK, DeLuca J, Diamond BJ, Tramontano G,Averill A. Speed of information processing in traumatic

e102 Neurology 73 November 24, 2009

brain injury: modality-specific factors. J Head Trauma Re-habil 2000;15:943–956.

16. Wills S, Leathem J. The effects of test anxiety, age, intelli-gence level, and arithmetic ability on Paced Auditory SerialAddition Test performance. Appl Neuropsychol 2004;11:180–187.

17. Bonnet MH, Arand DL. We are chronically sleep de-prived. Sleep 1995;18:908–911.

18. Dinges DF, Pack F, Williams K, et al. Cumulative sleepi-ness, mood disturbance, and psychomotor vigilance per-formance decrements during a week of sleep restricted to4–5 hours per night. Sleep 1997;20:267–277.

Neurology 73 November 24, 2009 e103

T h e n e w e ngl a nd j o u r na l o f m e dic i n e

n engl j med 361;14 nejm.org october 1, 2009 1349

original article

Moderate Hypothermia to Treat Perinatal Asphyxial Encephalopathy

Denis V. Azzopardi, F.R.C.P.C.H., Brenda Strohm, R.G.N., A. David Edwards, F.Med.Sci., Leigh Dyet, M.B., B.S., Ph.D.,

Henry L. Halliday, F.R.C.P.H., Edmund Juszczak, M.Sc., Olga Kapellou, M.D., Malcolm Levene, F.Med.Sci., Neil Marlow, F.Med.Sci.,

Emma Porter, M.R.C.P.C.H., Marianne Thoresen, M.D., Ph.D., Andrew Whitelaw, F.R.C.P.C.H., and Peter Brocklehurst, F.F.P.H.,

for the TOBY Study Group*

From the Division of Clinical Sciences and Medical Research Council (MRC) Clinical Sciences Centre, Hammersmith Hospital, Imperial College London, Lon-don (D.V.A., A.D.E., L.D., O.K., E.P.); the National Perinatal Epidemiology Unit, Uni-versity of Oxford, Oxford (B.S., E.J., P.B.); the Department of Perinatal Medicine, Royal Maternity Hospital and Depart-ment of Child Health, Queen’s University, Belfast (H.L.H.); the University of Leeds and Leeds General Infirmary, Leeds (M.L.); the Academic Division of Child Health, Queen’s Medical Centre, Nottingham (N.M.); and the Department of Clinical Science, University of Bristol, St. Michael’s Hospital (M.T.) and Southmead Hospital (A.W.), Bristol — all in the United King-dom. Address reprint requests to Dr. Az-zopardi at the Division of Clinical Scienc-es and MRC Clinical Sciences Centre, Imperial College London, Hammersmith Hospital, Du Cane Rd., London W12 0NN, United Kingdom, or at d.azzopardi@ imperial.ac.uk.

*The members of the Total Body Hypo-thermia for Neonatal Encephalopathy Trial (TOBY) study group are listed in the Ap-pendix.

N Engl J Med 2009;361:1349-58.Copyright © 2009 Massachusetts Medical Society.

A bs tr ac t

Background

Whether hypothermic therapy improves neurodevelopmental outcomes in newborn infants with asphyxial encephalopathy is uncertain.

Methods

We performed a randomized trial of infants who were less than 6 hours of age and had a gestational age of at least 36 weeks and perinatal asphyxial encephalopathy. We compared intensive care plus cooling of the body to 33.5°C for 72 hours and in-tensive care alone. The primary outcome was death or severe disability at 18 months of age. Prespecified secondary outcomes included 12 neurologic outcomes and 14 other adverse outcomes.

Results

Of 325 infants enrolled, 163 underwent intensive care with cooling, and 162 under-went intensive care alone. In the cooled group, 42 infants died and 32 survived but had severe neurodevelopmental disability, whereas in the noncooled group, 44 in-fants died and 42 had severe disability (relative risk for either outcome, 0.86; 95% confidence interval [CI], 0.68 to 1.07; P = 0.17). Infants in the cooled group had an increased rate of survival without neurologic abnormality (relative risk, 1.57; 95% CI, 1.16 to 2.12; P = 0.003). Among survivors, cooling resulted in reduced risks of cere-bral palsy (relative risk, 0.67; 95% CI, 0.47 to 0.96; P = 0.03) and improved scores on the Mental Developmental Index and Psychomotor Developmental Index of the Bayley Scales of Infant Development II (P = 0.03 for each) and the Gross Motor Function Classification System (P = 0.01). Improvements in other neurologic outcomes in the cooled group were not significant. Adverse events were mostly minor and not associ-ated with cooling.

Conclusions

Induction of moderate hypothermia for 72 hours in infants who had perinatal as-phyxia did not significantly reduce the combined rate of death or severe disability but resulted in improved neurologic outcomes in survivors. (Current Controlled Trials number, ISRCTN89547571.)

Copyright © 2009 Massachusetts Medical Society. All rights reserved. Downloaded from www.nejm.org at UNIVERSITY OF WASHINGTON on December 10, 2009 .


n engl j med 361;14 nejm.org october 1, 20091350

Perinatal asphyxial encephalopathy is associated with high morbidity and mor-tality rates worldwide and is a major burden

for the patient, the family, and society. There is an urgent need to improve outcomes in affected in-fants.

Experimentally, reducing body temperature to 3 to 5°C below the normal level reduces cerebral injury and improves neurologic function after as-phyxia.1-6 Preliminary clinical studies have found no serious adverse effects of cooling.7-9 Two ran-domized, controlled trials, the CoolCap trial10 and the National Institute of Child Health and Hu-man Development (NICHD) trial,11 have reported outcomes among infants at 18 months of age who had asphyxial encephalopathy, after slightly dif-ferent cooling regimens. Only the NICHD trial showed a significant reduction in the composite primary outcome of death or disability with hypo-thermia. Neither trial had sufficient power to de-tect significant differences in important individu-al neurologic outcomes, and several systematic reviews and an expert workshop did not reach a consensus in recommending hypothermia as stan-dard treatment.12-17

To clarify the role of hypothermia, we carried out the Total Body Hypothermia for Neonatal En-cephalopathy Trial (TOBY), a multicenter, random-ized trial comparing intensive care plus total-body cooling for 72 hours with intensive care without cooling among term infants with asphyxial en-cephalopathy.

Me thods

The TOBY protocol was approved by the London Multicenter Research Ethics Committee and the local research ethics committee of each participat-ing hospital. Conduct of the study was overseen by an independent trial steering committee with advice from an independent data monitoring and ethics committee.

Study Design and Procedures

TOBY was a randomized, controlled trial, involv-ing term infants, comparing intensive care plus total-body cooling for 72 hours with intensive care without cooling. Infants were eligible if they were born at or after 36 completed weeks’ gestation. They also had to have, at 10 minutes after birth, either an Apgar score of 5 or less or a continued need for resuscitation or, within 60 minutes after

birth, acidosis (defined as any occurrence of um-bilical-cord, arterial, or capillary pH of <7.00 or base deficit of ≥16 mmol per liter). In addition, they had to have moderate-to-severe encephalopa-thy (indicated by lethargy, stupor, or coma) and either hypotonia, abnormal reflexes (including oc-ulomotor or pupillary abnormalities), an absent or weak suck, or clinical seizures. Finally, they had to have abnormal background activity of at least 30 minutes’ duration or seizures on amplitude-integrated electroencephalography.18

We excluded infants expected to be more than 6 hours of age at the time of randomization and those with major congenital abnormalities known at randomization that required surgery or were suggestive of chromosomal anomaly or syndromes that involve brain dysgenesis.

Written informed consent was obtained from a parent of each infant after explanation of the study, and consent was reaffirmed within the sub-sequent 24 hours.19 Assignment to a treatment group was performed by means of central tele-phone randomization or a secure Web-based sys-tem (provided by the National Perinatal Epide-miology Unit Clinical Trials Unit, Oxford, United Kingdom). Minimization was used to ensure bal-ance of treatment assignment among infants with various grades of abnormality on amplitude-inte-grated electroencephalography and within each participating center.

Clinical Management

All recruited infants were cared for in partici-pating centers. Infants from referring hospitals were assessed by trained retrieval teams who per-formed amplitude-integrated electroencephalog-raphy, sought consent if the infant was eligible, performed randomization, and for infants assigned to the cooled group, began cooling by discontinu-ing warming and applying cooled gel packs, if nec-essary, until the infant was admitted to a partici-pating center.

To minimize potential confounding from dif-ferential use of cointerventions, uniform guidance was provided on ventilatory and circulatory care, management of seizures, sedation, and fluid re-quirements. All infants underwent sedation with morphine infusions or with chloral hydrate if they appeared to be distressed. Skin temperature and rectal temperature (measured at least 2 cm within the rectum) were monitored continuously and re-corded hourly in all infants throughout the inter-


Moder ate Hypothermia for Perinatal Asphyxial Encephalopathy


vention period. Clinical staff were made aware of the treatment assignments so that they could man-age cooling appropriately.

Intensive Care AloneInfants assigned to the noncooled group received the current standard of care and were placed un-der radiant heaters or in incubators, which were servo-controlled according to the abdominal skin temperature to maintain the rectal temperature at 37.0±0.2°C.

Intensive Care with CoolingInfants assigned to the cooled group were treated in incubators with the power turned off. Hypo-thermia was maintained by nursing the infant on a cooling blanket in which fluid whose tempera-ture was regulated by a manually adjusted thermo-stat (Tecotherm TS 200, Tec-Com) was circulated. The target rectal temperature was 33 to 34°C, and typically, the thermostat was set from 25 to 30°C.

Rewarming ProceduresWhen the period of cooling concluded, 72 hours after randomization, the rectal temperature was monitored for at least 4 hours to prevent rebound hyperthermia. The rectal temperature was allowed to rise by no more than 0.5°C per hour, to a max-imum of 37±0.2°C. Cranial ultrasonography was performed daily for the first 4 days after birth, and magnetic resonance imaging (MRI) was con-ducted, according to a specified protocol, within 5 to 14 days after birth.

Outcomes

The primary outcome at 18 months of age was a composite of death or severe neurodevelopmental disability in survivors. Severe neurodevelopmental disability was defined as a score of less than 70 on the Mental Developmental Index of the Bayley Scales of Infant Development II (BSID-II) (on which the standardization mean [±SD] is 100±15 and higher scores indicate better performance), a score of 3 to 5 on the Gross Motor Function Classifica-tion System20 (GMFCS) (on which scores can range from 1 to 5, with higher scores indicating greater impairment), or bilateral cortical visual impairment with no useful vision.

Adverse outcomes included intracranial hemor-rhage, persistent hypotension, pulmonary hemor-rhage, pulmonary hypertension, prolonged blood coagulation time, culture-proven sepsis, necrotiz-ing enterocolitis, cardiac arrhythmia, thrombocy-

topenia, major venous thrombosis, renal failure treated with dialysis, pneumonia, pulmonary air leak, and duration of hospitalization. (Most out-comes are defined in the Supplementary Appen-dix, available with the full text of this article at NEJM.org.)

Secondary outcomes at 18 months, specified before data analysis, included death and severe neurodevelopmental disability (components of the composite primary outcome), the score on the Psychomotor Developmental Index of BSID-II (on which the standardization mean [±SD] is 100±15 and higher scores indicate better performance), cerebral palsy, hearing loss, seizures treated with anticonvulsant agents, microcephaly (i.e., age- and sex-standardized head circumference of more than 2 SD below the mean), multiple neurodevelop-mental abnormalities (i.e., more than one of the following: a GMFCS score of 3 to 5, a score of <70 on the Mental Developmental Index of BSID-II) (on which the standardization mean [±SD] is 100±15 and higher scores indicate better performance, seizures, or cortical visual impairment and hear-ing loss), and survival without neurologic abnor-mality (i.e., a Mental Developmental Index score >84, a Psychomotor Developmental Index score >84, no abnormalities on GMFCS assessment, and normal vision and hearing).

Neurologic Assessment

Infants were assessed at approximately 18 months of age, through a structured examination by one of five trained assessors who were unaware of the treatment assignments. Neurologic signs and func-tion were scored,20,21 and the presence and type of cerebral palsy were determined. Neurodevel-opmental outcome was assessed according to the BSID-II.22

Statistical Analysis

We estimated that a sample of 236 infants would be required to detect a relative risk of 0.6 to 0.7 for the primary outcome in the cooled group as com-pared with the noncooled group, with a statistical power of 80%, at a two-sided significance level of 5% and assuming a 10% loss to follow-up. This sample size was achieved ahead of schedule, and enrollment was continued after the CoolCap and NICHD trial results suggested that a larger sam-ple would be valuable.

Demographic and clinical characteristics were summarized at baseline as counts and percentages of the total numbers of infants for categorical vari-




ables, as means (±SD) for normally distributed continuous variables, and as medians and ranges or interquartile ranges for other continuous vari-ables.

Data were analyzed in the groups to which pa-tients had been assigned regardless of either de-viation from the protocol or treatment received. Consistent with previous reports,10,11 neurologic outcomes are presented for survivors who had available follow-up data.

Comparative statistical analysis entailed calcu-lating the relative risks plus the 95% confidence intervals for all dichotomous outcomes, the mean differences plus 95% confidence intervals for nor-mally distributed, continuous outcomes (using analysis of covariance where appropriate), and the median differences plus 95% confidence intervals for skewed continuous variables. In addition, or-dered categorical variables were examined with the use of the chi-square test for trend. An adjusted analysis of the primary outcome was performed to investigate the effect of known prognostic fac-tors. All statistical tests were two-sided and were not adjusted for multiple comparisons.

Prespecified subgroup analyses were performed

with stratification on the basis of the grade of abnormality on amplitude-integrated electroen-cephalography at randomization and duration of the interval between birth and randomization (0 to <4 hours vs. 4 to 6 hours). The consistency of the effect of the treatment group across subgroups was explored by means of the statistical test of interaction.

R esult s

From December 1, 2002, through November 30, 2006, 494 infants were screened and 325 were recruited from 42 hospitals (Fig. 1). The infants were from the United Kingdom (277), Hungary (24), Sweden (18), Israel (4), and Finland (2). Base-line characteristics of the infants (including ma-ternal characteristics) were broadly similar between the two groups (Table 1).

Compliance with Cooling Protocol

Rectal temperatures were similar between the two groups at the time of randomization (Table 1). Mean rectal temperatures at 6 to 72 hours after randomization were 33.5±0.5°C and 36.9±0.6°C in the cooled and noncooled groups, respectively (Fig. 2). Among the 162 infants who were not cooled, during the treatment period the tempera-ture rose above 38°C on one occasion in 14 (9%) and on more than one occasion in 23 (14%).

Primary Outcome

In the cooled group, 42 infants died and 32 sur-vived with severe neurodevelopmental disability, whereas in the noncooled group, 44 infants died and 42 had severe disability (Table 2) (relative risk for either outcome, 0.86; 95% confidence interval [CI], 0.68 to 1.07; P = 0.17). The result was materi-ally unchanged when adjusted for severity of ab-normality on amplitude-integrated electroenceph-alography, sex, or age at randomization.

Adverse Outcomes

The incidence of adverse events was similar in the two groups (Table 3). Hypotension, thrombocy-topenia, prolonged coagulation time, and intra-cranial hemorrhage (seen only on MRI) were fre-quently observed in both groups. Serious adverse events other than death were uncommon and were not associated with cooling. Two infants in the cooled group and one in the noncooled group had sinus thrombosis noted on MRI. Another infant in the cooled group had a thrombus in the aorta,

22p3

325 Underwent randomization

494 Patients were assessed for eligibility

169 Were excluded94 Did not meet inclusion

criteria30 Declined to participate45 Had other reasons

163 Were assigned to undergointensive care plus cooling

162 Were assigned to undergointensive care without cooling

1 Was lost to follow-up (unableto collect outcome data)

42 Died120 Survivors were assessed

1 Was lost to follow-up (unableto collect outcome data)

44 Died117 Survivors were assessed

163 Were analyzed 162 Were analyzed

AUTHOR:

FIGURE:

JOB: ISSUE:

4-CH/T

RETAKE

SIZE

ICM

CASE

EMail LineH/TCombo

Revised

AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset.

Please check carefully.

REG F

Enon

1st

2nd3rd

Azzopardi

1 of 2

xx-xx-09

ARTIST: ts

360xx

Figure 1. Enrollment and Follow-up of the Study Infants.




as well as an umbilical arterial catheter and a he-matocrit of 70%. No case of renal failure requir-ing dialysis occurred.

Secondary Outcomes at 18 Months

The mortality rate was similar in both groups. Forty-two infants in the cooled group died, as did 44 infants in the noncooled group; in each group, 39 of those died before hospital discharge. Death

occurred after the withdrawal of care in 34 of the 39 (87%) in the cooled group and 29 of the 39 (74%) in the noncooled group.

Outcomes were significantly improved in the cooled group with regard to 5 of the 12 second-ary neurologic outcomes assessed (Table 2). The rate of survival without a neurologic abnormality was significantly increased in the cooled group (71 of 163 infants [44%], vs. 45 of 162 [28%] in

Table 1. Baseline Characteristics of the Infants.*

CharacteristicCooled Group

(N = 163) Noncooled Group

(N = 162) P Value

Male sex — no. (%) 101 (62) 88 (54) 0.16

Gestational age — wk

Median 40.3 40.1 0.22

IQR 39.1–41.3 38.8–41.1

Birth weight — g

Median 3450 3350 0.18

IQR 2957–3873 3044–3729

Head circumference — cm

Median 35.0 35.0 0.53

IQR 34.0–36.0 34.0–35.9

Age at randomization

Median — hr 4.7 4.7 0.88

IQR — hr 3.8–5.4 3.5–5.5

0 to <4 hr — no. (%) 48 (29) 57 (35) 0.27

4 to 6 hr — no. (%) 115 (71) 105 (65)

Maternal pyrexia during labor — no. (%)† 10 (6) 10 (6) 0.94

Delivery complications — no. (%) 115 (71) 119 (74) 0.55

Apgar score at 10 min

Median 4 4 0.15

IQR 2–5 2–5

<5 — no. (%) 110 (83) 105 (77) 0.21

Resuscitation required at 10 min of age — no. (%) 149 (91) 151 (93) 0.54

Clinical seizures — no. (%) 92 (56) 83 (51) 0.35

Temperature at randomization

Mean ±SD — °C 36.6±1.1 36.5±1.2 0.24

<35.5°C — no. (%) 18 (11) 25 (16) 0.24

Abnormality on aEEG at randomization — no. (%)

Moderate 65 (40) 67 (41)

Severe 98 (60) 95 (59) 0.79

* Data were unavailable for some patients: for gestational age, 16 patients in the cooled group and 17 in the noncooled group; for birth weight, 1 patient in the cooled group and 1 in the noncooled group; for head circumference, 41 pa-tients in the cooled group and 44 in the noncooled group; for maternal pyrexia, 2 patients in the cooled group and 6 in the noncooled group; for delivery complications, 2 patients in the cooled group and 2 in the noncooled group; for Apgar score at 10 minutes, 31 patients in the cooled group and 26 in the noncooled group; and for temperature at ran-domization, 2 patients in the noncooled group. The term aEEG denotes amplitude-integrated electroencephalography, and IQR interquartile range.

† Pyrexia was defined as a temperature of 37.6°C or more.




the noncooled group; relative risk, 1.57; 95% CI, 1.16 to 2.12; P = 0.003). Among survivors, cooling resulted in reduced risks of cerebral palsy (relative risk, 0.67; 95% CI, 0.47 to 0.96; P = 0.03) and ab-normal GMFCS score (relative risk, 0.63; 95% CI, 0.45 to 0.89; P = 0.007) and resulted in improve-ments in the Mental Developmental Index and Psychomotor Developmental Index scores (P = 0.03 for each), and the GMFCS score (P = 0.01). The rate of multiple neurodevelopmental abnormali-ties was 21 of 112 in the cooled group, as com-pared with 33 of 110 in the noncooled group (relative risk, 0.63; 95% CI, 0.39 to 1.01; P = 0.05). Results from the complete analysis of the neurode-velopmental assessments are provided in the Sup-plementary Appendix.

Subgroup Analyses

More infants with severely abnormal results on amplitude-integrated electroencephalography at the time of randomization died or had a severe disability than did those with moderately abnor-mal results (109 of 193 [56%] vs. 50 of 132 [38%]; relative risk, 1.49; 95% CI, 1.16 to 1.91; P = 0.001).

However, the effect of cooling did not significant-ly vary according to the severity of abnormality on amplitude-integrated electroencephalography (P = 0.23 for interaction). The results were similar when the analysis was carried out with the results of amplitude-integrated electroencephalography classified as in the CoolCap study.10 The effect of treatment group did not vary significantly on the basis of time to randomization: among the 105 infants randomly assigned to a group less than 4 hours after birth, the relative risk for the primary outcome with cooling was 0.77 (95% CI, 0.44 to 1.04), whereas among the 220 remaining infants who were randomly assigned between 4 and 6 hours after birth, the relative risk was 0.95 (95% CI, 0.72 to 1.25; P = 0.21 for interaction).

Discussion

In this trial of near-term infants with perinatal asphyxia, we found no significant difference in the risk of the primary outcome, the combined rates of death or severe disability, between the cooled group and the noncooled group. However, cooling resulted in consistent improvement in second-ary outcomes, including a significant increase in the rate of survival without neurologic abnormal-ities and improved neurodevelopmental outcomes among survivors.

The primary outcome of TOBY, as in the CoolCap and NICHD trials, was a composite end point, chosen because of concerns that cooling might increase survival with additional disability. Results of all three trials are consistent with re-spect to this primary outcome, with point esti-mates supporting a benefit from hypothermia: the relative risk associated with cooling (vs. no cool-ing) was 0.82 (95% CI, 0.66 to 1.02) in the CoolCap study, 0.72 (95% CI, 0.71 to 0.93) in the NICHD trial (which included infants with moderate dis-ability), and 0.86 (95% CI, 0.68 to 1.07) in the present trial.

Our categorization of neurologic outcomes is consistent with that used in the CoolCap and NICHD trials and previous systematic reviews,12-17 facilitating the comparison of our findings with previous results. We found a significant increase in the rate of survival without neurologic abnor-mality with cooling (relative risk, 1.57; 95% CI, 1.16 to 2.12); the NICHD and CoolCap trials, which were smaller than the present trial, showed nonsignificant benefits with regard to this out-

22p3

40

Mea

n R

ecta

l Tem

pera

ture

(°C

)

38

39

37

36

34

33

30

35

32

31

00 4 8 12 16 20 24 28 32 36 40 48 52 56 60 64 68 72 76 80 8444 88

Cooling

No cooling

Hours since Randomization

AUTHOR:

FIGURE:

JOB:

4-CH/T

RETAKE

SIZE

ICM

CASE

EMail LineH/TCombo

Revised

AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset.

Please check carefully.

REG F

Enon

1st

2nd3rd

Azzopardi

2 of 2

xx-xx-09

ARTIST: ts

360xx ISSUE:

Figure 2. Mean Rectal Temperatures during the Study, According to Treatment Group.

The vertical bars indicate 2 SDs.




come but had similar point estimates. The relative risk in the NICHD trial was 1.51 (95% CI, 0.94 to 2.42),23 and the relative risk in the CoolCap trial was 1.48 (95% CI, 0.89 to 2.45) (Gunn A, University of Auckland, New Zealand: personal communication).

Although there is striking homogeneity among results of these three trials, there are also some differences. Only the NICHD trial detected a sig-nificant effect of hypothermia on the primary outcome, and only TOBY detected significant im-provements in specific neurologic outcomes. Both

TOBY and the CoolCap trial showed that the in-creased risk of death or severe disability in infants with the most abnormal grade on amplitude-integrated electroencephalography was unaffect-ed by cooling, but the CoolCap results suggested a reduction of the risk in the subgroup with less severely abnormal findings.

These discrepancies among results may be ex-plained in part by differences in the trial proto-cols. In all three trials, the whole-body tempera-ture (as measured in the rectum or esophagus) was reduced, but the strategies for cooling and

Table 2. Main Neurodevelopmental Outcomes at 18 Months.

Outcome Cooled Group Noncooled Group P ValueRelative Risk

(95% CI)

no./total no. (%)

Primary outcome

Combined death and severe neurodevelopmental disability

74/163 (45) 86/162 (53) 0.17 0.86 (0.68–1.07)

Secondary outcomes*

Death 42/163 (26) 44/162 (27) 0.78 0.95 (0.66–1.36)

Severe neurodevelopmental disability 32/120 (27) 42/117 (36) 0.13 0.74 (0.51–1.09)

Survival without neurologic abnormality 71/163 (44) 45/162 (28) 0.003 1.57 (1.16–2.12)

Multiple neurodevelopmental disabilities 21/112 (19) 33/110 (30) 0.05 0.63 (0.39–1.01)

BSID-II Mental Developmental Index score 0.03 for trend

<70 28/115 (24) 38/110 (35) 0.09 0.70 (0.47–1.06)

70–84 6/115 (5) 12/110 (11)

≥85 81/115 (70) 60/110 (55) 0.01 1.29 (1.05–1.59)

BSID-II Psychomotor Developmental Index score

0.03 for trend

<70 27/114 (24) 37/109 (34) 0.09 0.70 (0.46–1.06)

70–84 9/114 (8) 14/109 (13)

≥85 78/114 (68) 58/109 (53) 0.02 1.29 (1.04–1.60)

GMFCS score 0.01 for trend

No abnormality 85/120 (71) 63/117 (54) 0.007 1.32 (1.07–1.61)

1–2 11/120 (9) 18/117 (15)

3–5 24/120 (20) 36/117 (31) 0.06 0.65 (0.41–1.02)

Cerebral palsy 33/120 (28) 48/117 (41) 0.03 0.67 (0.47–0.96)

Hearing loss not corrected by aids 4/114 (4) 7/108 (6) 0.31 0.54 (0.16–1.80)

No useful vision 8/119 (7) 12/114 (11) 0.30 0.64 (0.27–1.50)

Seizures requiring anticonvulsant agents at time of assessment

12/116 (10) 16/116 (14) 0.42 0.75 (0.37–1.51)

Head circumference at follow-up >2 SD below the mean

24/114 (21) 28/112 (25) 0.48 0.84 (0.52–1.36)

* Scores on the Bayley Scales of Infant Development II (BSID-II) are assessed relative to a standardization mean (±SD) of 100±15, with higher scores indicating better performance. Scores on the Gross Motor Function Classification System (GMFCS) can range from 1 to 5, with higher scores indicating greater impairment. CI denotes confidence interval.




the target temperatures varied: temperature was decreased to 33 to 34°C with the use of cooling blankets in TOBY and the NICHD trial and to 34 to 35°C by means of scalp cooling in the CoolCap study. The NICHD trial recorded slight-ly higher temperatures in the control group than did the other two trials. In TOBY, but not the other two trials, cooling was initiated during transport to the treatment center. In TOBY and the CoolCap trial, but not the NICHD trial, pa-tients were selected on the basis of the presence of abnormalities on amplitude-integrated electro-encephalography in addition to clinical criteria. Differences in local practices for withdrawal of care may also have affected outcomes. Withdrawal was slightly more common in the control group than in the cooled group in the NICHD study but was more common in the cooled group than in the control group in TOBY; these results may par-tially account for the greater apparent effect of

hypothermia on mortality rate in the NICHD study as compared with TOBY.

Elevation of body temperature to greater than 38°C was observed in several noncooled infants in the NICHD and CoolCap trials and was associ-ated with a worse outcome in the CoolCap trial.11,24 A rectal temperature of more than 38°C was also noted in some noncooled infants in TOBY. Experi-mental data showing that pyrexia may adversely affect neurodevelopment support the possibility that increased temperatures may contribute to the poorer outcomes seen in the noncooled groups; however, it is also possible that the relationship between higher elevation of body temperature and poor outcome reflects reverse causation (i.e., as-phyxia resulting in impairment of temperature regulation).

Consistent with findings in the earlier trials, in our study, minor respiratory and cardiovascular events were common, but serious adverse events

Table 3. Adverse Outcomes, According to Treatment Group.*

OutcomeCooled Group

(N = 163)Noncooled Group

(N = 162) P Value

Relative Risk or Median Difference

(95% CI)†

Total duration of hospital care — days

Median 12 13 0.13 1 (0–4)

IQR 8–18 9–25

Persistent hypotension — no./total no. (%)‡ 126/163 (77) 134/162 (83) 0.22 0.93 (0.84–1.04)

Prolonged coagulation time — no./total no. (%) 67/163 (41) 72/161 (45) 0.51 0.92 (0.71–1.18)

Thrombocytopenia — no./total no. (%) 94/163 (58) 80/161 (50) 0.15 1.16 (0.95–1.42)

Intracranial hemorrhage — no./total no. (%)§ 25/64 (39) 21/67 (31) 0.35 1.25 (0.78–1.99)

Pulmonary diagnoses — no./total no. (%)

Pneumonia 5/163 (3) 5/162 (3) 0.99 0.99 (0.29–3.37)

Pulmonary air leak 9/163 (6) 3/162 (2) 0.08 2.98 (0.82–10.80)

Pulmonary hemorrhage 5/163 (3) 3/162 (2) 0.48 1.66 (0.40–6.82)

Pulmonary hypertension 16/163 (10) 9/162 (6) 0.15 1.77 (0.80–3.88)

Necrotizing enterocolitis — no./total no. (%) 1/163 (<1) 0/162

Cardiac arrhythmia — no./total no. (%)¶ 8/163 (5) 3/162 (2) 0.13 2.65 (0.72–9.81)

Culture-proven sepsis — no./total no. (%) 20/163 (12) 20/162 (12) 0.98 0.99 (0.56–1.78)

* No case of renal failure requiring dialysis occurred. CI denotes confidence interval, and IQR interquartile range.† Values are relative risks except for total duration of hospital care, for which the median difference is shown.‡ Hypotension was defined as a mean blood pressure of 40 mm Hg or less.§ Intracranial hemorrhage was identified on magnetic resonance imaging (MRI); 39 of the 46 cases were subdural (10

moderate and 29 mild). No intracranial hemorrhage was identified on cranial ultrasonography. Two cases of sinus thrombosis in the cooling group, and one case in the noncooled group, were also noted on MRI.

¶ All but three cases of cardiac arrhythmia consisted of sinus bradycardia of less than 80 beats per minute. The other three cases consisted of ventricular arrhythmia; two of these occurred in the noncooled group.




were rare and were not associated with cooling. Mild-to-moderate intracranial hemorrhage that was not visible on cranial ultrasonography was frequently seen on MRI in both groups, and sinus thrombosis occurred very infrequently in both groups.

No trial, to our knowledge, has yet reported neurologic outcomes at ages older than 18 months. Neurodevelopmental assessments at 18 months may not reliably predict later outcomes.25 Although it is likely that severe neuromotor disability will have been correctly identified at 18 months, less severe impairments are not reliably assessable at this age. Assessment later in childhood (e.g., at 6 to 7 years of age) is necessary for accurate, com-prehensive evaluation of cognitive function, be-

havior and learning, fine motor development, at-tention, and psychosocial health.26

In conclusion, TOBY did not show a significant reduction in the combined rates of death and se-vere disability with cooling, as compared with no cooling, but did show a significant improvement in several secondary neurologic outcomes among survivors. Whether this improvement is main-tained in the longer term needs to be ascertained.

Supported by grants from the U.K. Medical Research Council and the U.K. Department of Health.

No potential conflict of interest relevant to this article was re-ported.

We thank the Imperial College Healthcare Biomedical Research Centre and Bliss, the Special Care Baby Charity, for their advice and support, as well as all the parents and infants who took part in the study.

AppendixMembers of the TOBY Study Group are as follows: Project Management Group — D. Azzopardi (chief investigator), P. Brocklehurst (chief investigator), A.D. Edwards (principal investigator), H. Halliday (principal investigator), M. Levene (principal investigator), M. Thoresen (principal investigator), A. Whitelaw (principal investigator), S. Ayers (National Perinatal Epidemiology Unit [NPEU] informa-tion technology coordinator), U. Bowler (NPEU Clinical Trials Unit [CTU] senior trials manager), M. Gallagher (NPEU data manager), E. Juszczak (NPEU CTU head of trials), C. Mulhall (NPEU TOBY study coordinator), B. Strohm (NPEU TOBY research nurse and study coordinator); Writing Committee — D. Azzopardi (chief investigator), P. Brocklehurst (chief investigator), A.D. Edwards (principal in-vestigator), E. Juszczak (NPEU CTU head of trials); Trial Steering Committee — N. McIntosh (chair), Child Life and Health, University of Edinburgh, Edinburgh, United Kingdom (UK); D. Azzopardi, Imperial College London, London; H. Baumer, Derriford Hospital, Plymouth, UK; P. Brocklehurst, NPEU, University of Oxford, Oxford, UK; C. Doré, Medical Research Council (MRC) CTU; D. Elbourne, London School of Hygiene and Tropical Medicine, London; R. Parnell, Scope Data Monitoring and Ethics Committee; R. Cooke (chair), Liverpool Women’s Hospital, University of Liverpool, Liverpool, UK; H. Davies, National Research Ethics Service; A. Johnson, Univer-sity of Oxford, Oxford, UK; S. Richmond, Sunderland District General Hospital, University of Newcastle, Newcastle, UK; P. Yudkin, Division of Public Health and Primary Health Care, University of Oxford, Oxford, UK; Trial Statisticians (NPEU) — S. Gates, Edmund Juszczak, M. Quigley; Trial Health Economists (NPEU) — O. Eddama, J. Henderson, S. Petrou; Clinical Research Fellow — O. Kapellou; Follow-up Pediatricians — L. Dyet, E. Porter, Imperial College London, London; G. Mero, Jósa András County Hospital, Nyíregyháza, Hungary; B. Vollmer, Karolinska Institutet, Stockholm; E. Goldstein, Soroka Medical Center, Beersheeva, Israel; Specialist Adviser — B. Hutchon, Royal Free Hospital, London; Cranial Ultrasonography Interpretation — C. Hagmann, University College London, London; Bayley Scales of Infant Development II (BSID-II) Training — S. Johnson, University of Nottingham, Nottingham, UK; MRI Evaluation — L. Ramenghi, University of Milan, Milan; M. Rutherford, MRC Clinical Sciences Centre, Imperial College London, London; Centers for Recruitment and Data Collection for MRI Evaluation (in descending order of no. of infants recruited, in parentheses) — Hammer-smith Hospital, London (54) — D. Azzopardi, A.D. Edwards, O. Kapellou, P. Corcoran; First Department of Pediatrics, Semmelweis University Hospital, Budapest, Hungary (24) — M. Szabó, A. Róka, E. Bodrogi; Homerton Hospital, London (20) — E. Maalouf, C. Harris; Southmead Hospital, Bristol, UK (20) — A. Whitelaw, S. Lamburne; University College Hospital, London (19) — N. Robertson, A. Kapetanakis; St. George’s Hospital, London (18) — K. Farrer, L. Kay-Smith; Royal Maternity Hospital, Belfast, UK (17) — H. Hal-liday, D. Sweet; Liverpool Women’s Hospital, Liverpool, UK (15) — M. Weindling, A.S. Burke; St. Michael’s Hospital, Bristol, UK (14) — M. Thoresen, J. Tooley, J. Kemp; Leicester Royal Infirmary, Leicester, UK (12) — A. Currie, M. Hubbard; Royal Sussex County Hos-pital, Brighton, UK (10) — P. Amess; Queen Silvia’s Hospital, Gothenburg, Sweden (9) — K. Thiringer, A. Flisberg; Leeds General Infirmary, Leeds, UK (8) — M. Levene, A. Harrop; Nottingham City Hospital, Nottingham, UK (8) — S. Watkin, D. Jayasinghe; John Radcliffe Hospital, Oxford, UK (7) — E. Adams; Karolinska Institutet, Stockholm (6) — C. Lothian, M. Blennow; Medway Maritime Hospital, Gillingham, UK (6) — S. Rahman, B. Jani, K. Vandertak; Luton and Dunstable Hospital, Luton, UK (5) — S. Skinner, Y. Mil-lar; Queen’s Medical Centre, Nottingham, UK (5) — N. Marlow, S. Wardle; Jessop Wing, Sheffield, UK (4) — M. Smith; Royal Victoria Infirmary, Newcastle, UK (4) — J. Berrington; Soroka Medical Center, Beersheva, Israel (4) — K. Marks; Bradford Royal Infirmary, Bradford, UK (3) — S. Chatfield; Heartlands Hospital, Birmingham, UK (3) — S. Rose; New Cross Hospital, Wolverhampton, UK (3) — B. Kumararatne, L. Greig; Norfolk and Norwich University Hospital, Norwich, UK (3) — P. Clarke; Lund University Hospital, Lund, Sweden (3) — V. Fellman; Wishaw General Hospital, Wishaw, UK (3) — R. Abara; City Hospital, Birmingham, UK (2) — D. Armstrong; Erinville Hospital–Cork University Hospital, Cork, Ireland (2) — D. Murray; Hospital for Children and Adolescents, Helsinki (2) — M. Metsaranta; Queen Mother’s Hospital, Glasgow, UK (2) — J. Simpson; Singleton Hospital, Swansea, UK (2) — J. Matthes; Southern General Hospital, Glasgow, UK (2) — P. MacDonald; University Hospital of Wales, Cardiff, UK (2) — S. Cherian; Princess Royal Ma-ternity Hospital, Glasgow, UK (1) — L. Jackson; Royal Cornwall Hospital, Truro, UK (1) — P. Munyard; Royal Devon and Exeter Foundation Trust, Exeter, UK (1) — M. Quinn; St. Mary’s Hospital, Manchester, UK (1) — S. Mitchell; James Cook University Hospital, Middlesbrough, UK (0) — S. Sinha; Derriford Hospital, Plymouth, UK (0) — J. Eason; Department of Pediatrics, Oulu University Hos-pital, Oulu, Finland (0) — M. Hallman.




References

Thoresen M, Penrice J, Lorek A, et al. 1. Mild hypothermia after severe transient hypoxia-ischemia ameliorates delayed ce-rebral energy failure in the newborn pig-let. Pediatr Res 1995;37:667-70.

Sirimanne ES, Blumberg RM, Bossa-2. no D, et al. The effect of prolonged modi-fication of cerebral temperature on outcome after hypoxic-ischemic brain injury in the infant rat. Pediatr Res 1996;39:591-7.

Amess PN, Penrice J, Cady EB, et al. 3. Mild hypothermia after severe transient hypoxia-ischemia reduces the delayed rise in cerebral lactate in the newborn piglet. Pediatr Res 1997;41:803-8.

Edwards AD, Yue X, Squier MV, et al. 4. Specific inhibition of apoptosis after cere-bral hypoxia-ischaemia by moderate post-insult hypothermia. Biochem Biophys Res Commun 1995;217:1193-9.

Bona E, Hagberg H, Loberg EM, Bå-5. genholm R, Thoresen M. Protective effects of moderate hypothermia after neonatal hypoxia-ischemia: short- and long-term outcome. Pediatr Res 1998;43:738-45.

Colbourne F, Corbett D, Zhao Z, Yang 6. J, Buchan AM. Prolonged but delayed post-ischemic hypothermia: a long-term out-come study in the rat middle cerebral ar-tery occlusion model. J Cereb Blood Flow Metab 2000;20:1702-8.

Gunn AJ, Gluckman PD, Gunn TR. 7. Selective head cooling in newborn infants after perinatal asphyxia: a safety study. Pediatrics 1998;102:885-92.

Azzopardi D, Robertson NJ, Cowan 8. FM, Rutherford MA, Rampling M, Edwards AD. Pilot study of treatment with whole body hypothermia for neonatal encepha-lopathy. Pediatrics 2000;106:684-94.

Eicher DJ, Wagner CL, Katikaneni LP, 9.

et al. Moderate hypothermia in neonatal encephalopathy: safety outcomes. Pediatr Neurol 2005;32:18-24.

Gluckman PD, Wyatt JS, Azzopardi D, 10. et al. Selective head cooling with mild sys-temic hypothermia after neonatal enceph-alopathy: multicentre randomised trial. Lancet 2005;365:663-70.

Shankaran S, Laptook AR, Ehrenkranz 11. RA, et al. Whole-body hypothermia for neonates with hypoxic-ischemic enceph-alopathy. N Engl J Med 2005;353:1574-84.

Shah PS, Ohlsson A, Perlman M. Hy-12. pothermia to treat neonatal hypoxic ische-mic encephalopathy: systematic review. Arch Pediatr Adolesc Med 2007;161:951-8.

Schulzke SM, Rao S, Patole SK. A sys-13. tematic review of cooling for neuropro-tection in neonates with hypoxic ischemic encephalopathy — are we there yet? BMC Pediatr 2007;7:30.

Jacobs S, Hunt R, Tarnow-Mordi W, 14. Inder T, Davis P. Cooling for newborns with hypoxic ischaemic encephalopathy. Cochrane Database Syst Rev 2007;4: CD003311.

Higgins RD, Raju TN, Perlman J, et al. 15. Hypothermia and perinatal asphyxia: ex-ecutive summary of the National Institute of Child Health and Human Development workshop. J Pediatr 2006;148:170-5.

Barks JD. Current controversies in hy-16. pothermic neuroprotection. Semin Fetal Neonatal Med 2008;13:30-4.

Kirpalani H, Barks J, Thorlund K, 17. Guyatt G. Cooling for neonatal hypoxic ischemic encephalopathy: do we have the answer? Pediatrics 2007;120:1126-30.

al Naqeeb N, Edwards AD, Cowan FM, 18. Azzopardi D. Assessment of neonatal en-cephalopathy by amplitude-integrated elec-

troencephalography. Pediatrics 1999;103: 1263-71.

Allmark P, Mason S. Improving the 19. quality of consent to randomised con-trolled trials by using continuous consent and clinician training in the consent pro-cess. J Med Ethics 2006;32:439-43.

Palisano RJ, Hanna SE, Rosenbaum 20. PL, et al. Validation of a model of gross motor function for children with cerebral palsy. Phys Ther 2000;80:974-85.

Haataja L, Mercuri E, Regev R, et al. 21. Optimality score for the neurologic exami-nation of the infant at 12 and 18 months of age. J Pediatr 1999;135:153-61.

Bayley N. Bayley scales of infant devel-22. opment. 2nd ed. San Antonio, TX: Psy-chological Corporation, 1993.

Shankaran S, Pappas A, Laptook AR, 23. et al. Outcomes of safety and effective-ness in a multicenter randomized, con-trolled trial of whole-body hypothermia for neonatal hypoxic-ischemic encepha-lopathy. Pediatrics 2008;122(4):e791-e798.

Wyatt JS, Gluckman PD, Liu PY, et al. 24. Determinants of outcomes after head cooling for neonatal encephalopathy. Pe-diatrics 2007;119:912-21.

Barnett AL, Guzzetta A, Mercuri E, et 25. al. Can the Griffiths scales predict neuro-motor and perceptual-motor impairment in term infants with neonatal encepha-lopathy? Arch Dis Child 2004;89:637-43.

Voss W, Neubauer AP, Wachtendorf M, 26. Verhey JF, Kattner E. Neurodevelopmental outcome in extremely low birth weight in-fants: what is the minimum age for reliable developmental prognosis? Acta Paediatr 2007;96:342-7.Copyright © 2009 Massachusetts Medical Society.

powerpoint slides of journal figures and tables

At the Journal’s Web site, subscribers can automatically create PowerPoint slides. In a figure or table in the full-text version of any article at NEJM.org, click on Get PowerPoint Slide. A PowerPoint slide containing the image, with its title and reference citation, can then be downloaded and saved.


Articles

www.thelancet.com Vol 373 March 28, 2009 1105

Long-term risk of epilepsy after traumatic brain injury in children and young adults: a population-based cohort studyJakob Christensen, Marianne G Pedersen, Carsten B Pedersen, Per Sidenius, Jørn Olsen, Mogens Vestergaard

SummaryBackground The risk of epilepsy shortly after traumatic brain injury is high, but how long this high risk lasts is unknown. We aimed to assess the risk of epilepsy up to 10 years or longer after traumatic brain injury, taking into account sex, age, severity, and family history.

Methods We identifi ed 1 605 216 people born in Denmark (1977–2002) from the Civil Registration System. We obtained information on traumatic brain injury and epilepsy from the National Hospital Register and estimated relative risks (RR) with Poisson analyses.

Findings Risk of epilepsy was increased after a mild brain injury (RR 2·22, 95% CI 2·07–2·38), severe brain injury (7·40, 6·16–8·89), and skull fracture (2·17, 1·73–2·71). The risk was increased more than 10 years after mild brain injury (1·51, 1·24–1·85), severe brain injury (4·29, 2·04–9·00), and skull fracture (2·06, 1·37–3·11). RR increased with age at mild and severe injury and was especially high among people older than 15 years of age with mild (3·51, 2·90–4·26) and severe (12·24, 8·52–17·57) injury. The risk was slightly higher in women (2·49, 2·25–2·76) than in men (2·01, 1·83–2·22). Patients with a family history of epilepsy had a notably high risk of epilepsy after mild (5·75, 4·56–7·27) and severe brain injury (10·09, 4·20–24·26).

Interpretation The longlasting high risk of epilepsy after brain injury might provide a window for prevention of post-traumatic epilepsy.

Funding Danish Research Agency, P A Messerschmidt and Wife’s Foundation, Mrs Grethe Bønnelycke’s Foundation.

IntroductionTraumatic brain injury raises the risk of epilepsy,1 but little is known about the duration of the increased risk and the factors that modify the risk, especially in children and young adults.2 In hospital-based case series, the risk of epilepsy 1–2 years after moderate to severe brain injury is related to some CT or MRI fi ndings and is high in people who had had neurosurgical procedures.2–6 In a population-based study, age in people who had traumatic brain injury at age 65 years or older and time since and severity of injury were signifi cant risk factors for epilepsy,1 but only a few studies have included children and young adults.1–4 In some of these studies,2,4 acute seizures in the fi rst week after brain injury were associated with a high risk of epilepsy. Studies of epilepsy related to level of consciousness (eg, assessed with the Glasgow coma scale) and duration of post-traumatic amnesia after brain injury have given confl icting results.1,3,4

No eff ective prophylaxis for epilepsy after traumatic brain injury is available, and trials with preventive drug have been discouraging.7 However, better information about prognostic factors might help the development of new prevention strategies and treatment.5

We studied the risk of epilepsy in a large population-based cohort of children and young adults and considered time since injury, sex, age, severity, and family history of epilepsy.

MethodsStudy populationWe used data from the Danish Civil Registration System (CRS)8 to identify all people born in Denmark between Jan 1, 1977, and Dec 31, 2002. All liveborn children and new residents in Denmark are assigned a unique personal identifi cation number (CRS number) together with information on vital status, emigration from Denmark, and CRS numbers of mothers, fathers, and siblings. The CRS number links individual information in all national registries and provides identifi cation of family members and links parents with their children.

Identity of individuals in the study was blinded to the investigators, and the study did not involve contact with individual patients. The study therefore did not need approval from the ethics committee according to Danish laws, but the project was approved by the Danish Data Protection Agency.

Data collectionInformation about brain injury and epilepsy was obtained from the Danish National Hospital Register,9 which contains information on all discharges from Danish hospitals since 1977; outpatients have been included in the register since 1995. All treatment is free of charge for Danish residents. Patients admitted to the only private epilepsy hospital in Denmark are also recorded in the

Lancet 2009; 373: 1105–10

Published OnlineFebruary 23, 2009DOI:10.1016/S0140-6736(09)60214-2

See Comment page 1060

Department of Neurology, Aarhus University Hospital, Aarhus, Denmark (J Christensen MD, P Sidenius MD); Department of Clinical Pharmacology (J Christensen) and National Centre for Register-based Research (M G Pedersen MSc, C B Pedersen MSc), University of Aarhus, Denmark; Southern California Injury Prevention Research Centre (SCIPRC), School of Public Health, UCLA, CA, USA (J Olsen MD), Department of Epidemiology (J Olsen) and Department of General Practice (M Vestergaard MD), Institute of Public Health, University of Aarhus, Aarhus, Denmark

Correspondence to:Dr Jakob Christensen, Department of Neurology, Aarhus University Hospital, Norrebrogade 44, DK-8000 Aarhus C, [email protected]

Articles

1106 www.thelancet.com Vol 373 March 28, 2009

Danish National Hospital Register. Specialists in neurology working in private outpatient clinics also treat patients with epilepsy, but these contacts are not recorded in the Danish National Hospital Register.

Diagnostic information in the National Hospital Register is based on the International Classifi cation of Diseases, 8th revision (ICD-8) from 1977–93, and ICD-10 from 1994–2002.

Cohort members, their parents and siblings were classifi ed with epilepsy if they had been hospitalised or in outpatient care with a diagnosis of epilepsy (ICD-8: 345; ICD-10: G40,G41).10–13 By use of the CRS numbers, we linked parents and siblings registered with an epilepsy diagnosis in the National Hospital Register. A person was recorded as having a family history of epilepsy if the date of fi rst epilepsy diagnosis in a parent or sibling preceded their epilepsy diagnosis.

Cohort members were classifi ed with mild brain injury (concussion: ICD-8 850.99; ICD-10 S06.0), severe brain injury (structural brain injury: ICD-8 851.29-854.99; ICD-10 S06.1-S06.9), or skull fracture (ICD-8 800.99–801.09, 803.99; ICD-10: S02–S02.1, S02.7, S02.9), respectively, if they had been admitted or been in outpatient care with the relevant diagnosis.14,15 Time of onset of epilepsy and brain injury was defi ned as the fi rst day of the fi rst contact to the hospital with the relevant diagnosis.

The defi nition of mild brain injury (concussion) in Denmark is based on the defi nition given by the American Congress of Rehabilitation Medicine.16 The diagnostic criteria include a relevant direct trauma against the head manifesting with changed brain function (ie, loss of consciousness, amnesia, confusion/disorientation, or focal [temporary] neurological defi cit). Severity of mild brain injury should not include loss of consciousness longer than 30 min, a Glasgow coma scale of 13 or less after 30 min, or post-traumatic amnesia longer than 24 h.17 Severe brain injury (structural brain injury) includes brain contusion or intracranial haemorrhage. Skull fracture liable to be associated with disruption of brain function can occur alone or be associated with other types of brain injury and usually requires verifi cation with a radiograph or CT. Brain injuries recorded in the same patient within 14 days were categorised as the same event according to the hierarchy of brain injury—severe brain injury, skull fracture, and mild brain injury. For each type of brain injury, we calculated the age at fi rst brain injury (0–5, 5–10, 10–15, and ≥15 years), the length of fi rst admission (0, 1–6, 7–13, 14–27, and ≥28 days), and time since fi rst brain injury (0–6 months, 6 months to 1 year, 1–2, 2–3, 3–5, 5–10, and ≥10 years).

Statistical analysesPeople were followed from birth until onset of epilepsy, death, emigration from Denmark, or Dec 31, 2002, whichever came fi rst. The incidence rate ratio (for these analyses a good approximation of the relative risk, the term used in this Article) of epilepsy was estimated by

Patients diagnosed with epilepsy

New cases (per 1000 person-years)

Adjusted relative risk (95% CI)

p value

Time (years) since mild brain injury

0·0–0·5 162 4·67 5·46 (4·67–6·37) <0·0001

0·5–1·0 78 2·37 2·91 (2·33–3·64) <0·0001

1·0–2·0 109 1·78 2·26 (1·87–2·73) <0·0001

2·0–3·0 99 1·79 2·33 (1·91–2·84) <0·0001

3·0–5·0 138 1·50 1·99 (1·68–2·36) <0·0001

5·0–10·0 154 1·14 1·56 (1·33–1·83) <0·0001

≥10·0 97 1·00 1·51 (1·24–1·85) <0·0001

No mild injury 16 633 0·87 1·00 ··

Time (years) since severe brain injury

0·0–0·5 35 19·62 21·26 (15·25 to 29·62) <0·0001

0·5–1·0 19 11·52 13·45 (8·57 to 21·09) <0·0001

1·0–2·0 18 6·06 7·42 (4·68 to 11·79) <0·0001

2·0–3·0 11 4·26 5·40 (2·99 to 9·76) <0·0001

3·0–5·0 11 2·69 3·52 (1·95 to 6·35) <0·0001

5·0–10·0 15 3·22 4·40 (2·65 to 7·30) <0·0001

≥10·0 7 2·94 4·29 (2·04 to 9·00) 0·0001

No severe injury 17 354 0·89 1·00 ··

Time (years) since skull fracture

0·0–0·5 6 2·90 2·96 (1·33 to 6·60) 0·0078

0·5–1·0 6 2·99 3·51 (1·58 to 7·83) 0·0021

1·0–2·0 13 3·38 4·30 (2·50 to 7·41) <0·0001

2·0–3·0 5 1·39 1·81 (0·75 to 4·35) 0·1845

3·0–5·0 9 1·36 1·78 (0·93 to 3·42) 0·0838

5·0–10·0 16 1·21 1·55 (0·95 to 2·54) 0·0781

≥10·0 23 1·46 2·06 (1·37 to 3·11) 0·0005

No fracture 17 392 0·89 1·00 ··

Each form of injury led to a signifi cant (p<0·0001) increase in risk of epilepsy relative to people without brain injury. Relative risk (RR) was adjusted for age and interaction with sex and calendar year. RR of epilepsy in people with brain injury was modifi ed by time since fi rst admission with brain injury for mild (p<0·0001) and severe (p<0·0001) brain injury but not skull fracture (p=0·16).

Table 1: Time since fi rst admission with brain injury and relative risk (RR) of epilepsy

35

30

25

20

15

10

5

0

Rela

tive

risk o

f epi

leps

y

0 1 2 3 4 5 6 7 8 9 ≥10Years after injury

Mild brain injurySevere brain injurySkull fractureReference

Figure: Relative risk of epilepsy after brain injury in Denmark (1977–2002)

Articles


log-linear Poisson regression18 with the GENMOD procedure in SAS (version 8.1). Because incidence of epilepsy depends on age, sex, and calendar year,10 all the relative risks were adjusted for these factors. Age, calendar year, age at fi rst brain injury, duration of fi rst admission with brain injury, time since fi rst brain injury, and history of epilepsy in a parent or sibling were time dependent variables;19 all other variables were treated as time independent. Age was categorised in quarter year age levels from birth to the fi rst birthday, in 1 year age levels from the fi rst birthday to the 20th birthday, and as 20–21 years and ≥22 years. Calendar year was categorised in 1 year periods from 1977 to 2002. Likelihood ratio tests were used to calculate p values and 95% CIs were calculated by use of Wald’s test.19 The adjusted-score test20 suggested that the regression models were not subject to overdispersion.

Role of the funding sourceThe sponsors had no role in the study design, data collection, data analysis, data interpretation, or writing of the Article. All authors had full access to the data and approved the decision to submit the Article for publication in The Lancet.

ResultsWe followed-up 1 605 216 for a total of 19 527 337 person-years. During this study period, 78 572 people had at least one traumatic brain injury, and in the same period, 17 470 people developed epilepsy, of whom 1017 had a preceding brain injury. Follow-up was stopped before the end of the study period for 45 677 people (2·9%) because of emigration from Denmark (30 362 [1·9%]) or death (15 315 [1·0%]).

Relative to no brain injury, the risk of epilepsy was two times higher after mild brain injury (RR 2·22, 95% CI 2·07–2·38); seven times higher after severe brain injury (7·40, 6·16–8·89); and two-times higher after skull fracture (2·17, 1·73–2·71).

Tables 1–3 show the risk of epilepsy after brain injury according to time since fi rst admission with brain injury, age at fi rst brain injury, and duration of fi rst hospital stay with brain injury.

The risk of epilepsy after mild (p<0·0001) and severe (p<0·0001) brain injury was highest during the fi rst years after injury, but remained high for more than 10 years after the injury as compared with people without such a history (table 1, fi gure). For patients with skull fractures, risk of epilepsy did not vary signifi cantly with time since injury (p=0·16; table 1).

Brain injury was associated with an increased risk of epilepsy in all age groups (table 2). The risk increased with age for mild (p<0·0001) and severe (p=0·02) brain injury and was highest among people older than 15 years at injury.

Patients who had a long duration of hospital stay with severe brain injury (p<0·0001) and skull fracture

(p=0·02) had a notably high risk of epilepsy (table 3). For people with mild brain injury there was no association between duration of hospital stay and risk of epilepsy (p=0·73; table 3).

Table 4 shows the relative risk of epilepsy after brain injuries subdivided by family history of epilepsy. The relative risk of epilepsy with a family history of the disorder and mild brain injury is between what would have been predicted from a multiplicative model (3·37×2·24=7·54) and from an additive model (3·37+2·24–1=4·61; table 4). The relative risk estimate associated with severe brain injury and family history of epilepsy of is almost the same as would have been predicted from an additive model (3·35+7·81–1=10·16; table 4). We had very few people with epilepsy with skull fracture and a family history of epilepsy (table 4).

The relative risk of epilepsy after mild brain injury was higher among women (2·49, 2·25–2·76) than among men (2·01, 1·83–2·22; p=0·003). There was no interaction with sex for patients with skull fractures (p=0·59) or severe brain injury (0·22). We calculated the risk of epilepsy for patients registered with brain injury according to ICD-8 and ICD-10 (ie, patients diagnosed in the time period 1977 to 1993 and 1994 to 2002, respectively). For patients with mild brain injury, the risk of epilepsy was lower in the ICD-8 period (RR 1·89, 1·71–2·10) than in the ICD-10 period (2·61, 2·37–2·87; p<0·0001). For severe brain injury, the risk of epilepsy was almost the same in the ICD-8 period (7·17,

Number of patients with epilepsy*



p value

Age (years) at mild brain injury

0–5 365 1·64 2·06 (1·86–2·29) <0·0001

5–10 243 1·56 2·12 (1·87–2·41) <0·0001

10–15 117 1·54 2·25 (1·88–2·71) <0·0001

≥15 112 2·03 3·51 (2·90–4·26) <0·0001

No mild injury 16 633 0·87 1·00 ··

Age (years) at severe brain injury

0–5 51 6·26 7·20 (5·47–9·48) <0·0001

5–10 24 4·96 6·18 (4·14–9·23) <0·0001

10–15 11 3·56 4·91 (2·72–8·87) <0·0001

≥15 years 30 7·47 12·24 (8·52–17·57) <0·0001

No severe injury 17 354 0·89 1·00 ··

Age (years) at skull fracture

0–5 52 1·53 1·95 (1·49–2·56) <0·0001

5–10 17 2·12 2·86 (1·78–4·60) <0·0001

10–15 5 1·81 2·55 (1·06–6·12) 0·0368

≥15 years 4 1·71 2·75(1·03–7·34) 0·0433

No skull fracture 17 392 0·89 1·00 ··

Each form of injury led to a signifi cant (p<0·0001) increase in risk of epilepsy relative to people without brain injury. Relative risk (RR) was adjusted for age and interaction with sex and calendar year. RR of epilepsy in people with brain injury was modifi ed by age at fi rst admission with brain injury for mild (p<0·0001) and severe (p=0·02) brain injury but not skull fracture (p=0·55).

Table 2: Age at fi rst admission with brain injury and relative risk of epilepsy

Articles


5·19–9·91) and the ICD-10 period (7·51, 6·02–9·38; p=0·82). For skull fracture, the risks of epilepsy were comparable in the ICD-8 period (2·00, 1·54–2·58) and ICD-10 period (2·87, 1·85–4·46; p=0·17).

DiscussionAs previously shown in studies smaller than ours,1,2,6,21 risk of epilepsy increased after brain injury in relation to severity of brain injury. Risk was high for more than 10 years after the brain injuries even for mild brain injury (concussion), a fi nding in contrast to that of a previous study showing no increased risk of epilepsy 5 years after a mild brain injury.1 The discrepancy might result from diff erent inclusion criteria for mild brain injury and epilepsy,1,11 and an insuffi cient sample size to identify a moderate increase in risk.1 Our results suggest that time from brain injury to clinically overt symptoms (seizures) can span several years, leaving room for clinical intervention.5 However, animal studies suggest that a specifi c time window exists shortly after injury in which appropriate drugs might stop the epileptogenic process,22 and antiepileptogenic trials after brain injury in human beings have not shown drug treatment to be eff ective.7 In Denmark, seizure prophylaxis with antiepileptic drugs after brain injury was not used routinely in the study period.23

We defi ned the onset of epilepsy as the fi rst day of the fi rst contact, although this is only an approximation. There may be a delay from the fi rst seizure to diagnosis of epilepsy. We have previously validated the epilepsy diagnosis in a sample from the Danish National Hospital Register and found that 64% were registered in the Danish National Hospital Register within 1 year of fi rst seizure, and 90% were registered within 5 years.11 Diagnostic delay might, therefore, explain part of the increased risk of epilepsy after the brain injury. Likewise, a delay between brain injury and diagnosis (eg, in patients with chronic subdural haematoma), could bias the estimates of epilepsy shortly after a brain injury diagnosis, but this eff ect is likely to be small, especially in children.

Brain injury might be the fi rst presentation of epilepsy, in which the patient has a head trauma during an unwitnessed seizure (reverse causation). In a subanalysis, we excluded patients diagnosed with epilepsy within the fi rst 6 weeks of fi rst brain injury diagnosis and found that the high risk of epilepsy remained for all types of brain injury, albeit in an attenuated form (data not shown). Although patients with infrequent seizures might remain undiagnosed more than 6 weeks, this problem probably aff ects only a small part of the delayed association between brain injury and epilepsy. The risk of epilepsy increased slightly with age at time of mild brain injury, and was highest for people over 15 years of age, indicating that susceptibility to epilepsy after brain injury increases with age. This fi nding is in line with results of a previous study identifying people aged 65 or more as being at high risk of epilepsy after brain injury.1 Alternatively, the severity of brain injuries might increase with age, or doctors might be more likely to hospitalise younger children with less severe brain injuries, resulting in a

Number of patients with epilepsy


Adjusted relative risk# (95% CI)

p value

Hospital stay (days) for mild brain injury

0 256 1·73 2·22 (1·96–2·51) <0·0001

1–6 563 1·60 2·20 (2·02–2·40) <0·0001

7–13 9 2·08 3·01 (1·56–5·78) 0·0010

14–27 4 2·54 3·68 (1·38–9·82) 0·0091

≥28 5 2·05 2·94 (1·22–7·07) 0·0159

No mild injury 16 633 0·87 1·00 ··

Hospital stay (days) for severe brain injury

0 12 1·73 2·09 (1·19–3·68) <0·0108

1–6 24 3·71 4·82 (3·23–7·20) <0·0001

7–13 15 7·04 9·42 (5·68–15·63) <0·0001

14–27 18 13·11 18·01 (11·34–28·60) <0·0001

≥28 47 14·86 20·07 (15·06–26·74) <0·0001

No severe injury 17 354 0·89 1·00 ··

Hospital stay (days) for skull fracture

0 8 2·13 2·72 (1·36–5·45) 0·0046

1–6 48 1·35 1·77 (1·33–2·35) <0·0001

7–13 10 1·99 2·70 (1·45–5·03) 0·0017

14–27 4 3·08 4·01 (1·51–10·69) 0·0055

≥28 8 5·59 6·69 (3·35–13·38) <0·0001

No fracture 17 392 0·89 1·00 ··

Each form of injury led to a signifi cant (p<0·0001) increase in risk of epilepsy relative to people without brain injury. Relative risk (RR) was adjusted for age and interaction with sex and calendar year. RR of epilepsy in people with brain injury was modifi ed by duration of fi rst hospital stay with brain injury for severe brain injury (p<0·0001) and skull fracture (p=0·02) but not mild brain injury (p=0·73).

Table 3: Duration of fi rst hospital stay with brain injury and relative risk of epilepsy

No family history of epilepsy Family history of epilepsy

Number of patients with epilepsy


p value Number of patients with epilepsy


p value

Mild brain injury

No 15 511 1·00 ·· 1122 3·37 (3·17–3·58) <0·0001

Yes 766 2·24 (2·08–2·41) <0·0001 71 5·75 (4·56–7·27) <0·0001

Severe brain injury

No 16 166 1·00 ·· 1188 3·35 (3·16–3·56) <0·0001

Yes 11 7·81 (6·48–9·42) <0·0001 5 10·09 (4·20–24·26) <0·0001

Skull fracture

No 16 202 1·00 ·· 1190 3·35 (3·16–3·55) <0·0001

Yes 75 2·28 (1·81–2·86) <0·0001 3 2·71 (0·87–8·41) 0·0842

Any brain injury

No 15 338 1·00 ·· 1115 3·39 (3·19–3·61) <0·0001

Yes 939 2·47 (2·31–2·65) <0·0001 78 5·73 (4·58–7·16) <0·0001

Patients might have been exposed to more than one type of brain injury at separate admissions/outpatient visits. Relative risk adjusted for age and its interaction with sex and calendar year.

Table 4: Family history and relative risk of epilepsy after traumatic brain injury

Articles


low relative risk of post-traumatic epilepsy in young age groups.

Post-traumatic epilepsy is thought to be typical of symptomatic epilepsy (ie, determined by environmental factors). However, twin studies suggest that genetic factors also play a part in localisation-related epilepsies, most of which are thought to be symptomatic or probably symptomatic.24 Family history of epilepsy and mild brain injury independently contribute to the risk of epilepsy.25 Thus, people genetically predisposed to epilepsy (ie, with a family history of epilepsy) have a higher risk of epilepsy than do people without genetic predisposition when exposed to mild brain injury. To our knowledge, no previous studies have studied the risk of epilepsy after brain injury in fi rst degree relatives to patients with epilepsy. In animals, variation in the susceptibility of various rat strains to post-traumatic epilepsy might lend some support to the hypothesis of an underlying genetically determined tendency to develop post-traumatic epilepsy.26

Our registration of family history is not complete because some parents and older siblings might have been diagnosed before the Danish National Hospital Register was established (Jan 1, 1977). This mis-classifi cation is likely to cause an under estimation of the eff ect of family history on the risk of epilepsy.

The relative risk of epilepsy after mild brain injury was slightly higher in female than in male patients perhaps because female patients with epilepsy are more likely registered in the National Hospital Register because of sex-specifi c factors, such as pregnancy. Alternatively female brains might be more susceptible to epilepsy after mild brain injury than are male brain, as supported by a previous study showing that localisation-related epilepsy with no apparent structural cause is more prevalent in women than in men.27 The sex diff erence was not present for the other types of brain injury, suggesting that other mechanisms might be involved in post-traumatic epilepsy after skull fracture and more severe brain injuries.

The length of fi rst hospital stay with brain injury was associated with an increased risk of epilepsy for severe brain injury and cranial fractures. The length of admission is probably related to severity of brain injury.

Despite the length and completeness of follow up, the size of the study cohort, and the population-based nature of the study,8 we had limited clinical information. In a recent study, we validated the epilepsy diagnosis in the Danish National Hospital Register.11 We found a positive predictive value of an ICD-8 or ICD-10 epilepsy diagnosis according to ILAE criteria28 of 81% for epilepsy and 89% for single seizures, but identifi ed no epilepsy diagnoses based on acute symptomatic seizures.11 Thus, some patients registered with epilepsy in the present study do not fulfi l the diagnostic criteria, but the misclassifi cation would only bias the results of the present study away from the null hypothesis if the quality of the epilepsy registration diff ers between

patients with and without brain injury, which we fi nd unlikely.

The Danish Hospital Register does not capture all patients with epilepsy, because some outpatients might be treated in private practice. However, estimates of incidence (68·8 per 100 000 people per year) and prevalence (0·6%) of epilepsy in Denmark based on data from the Danish National Hospital Register were similar to those found in other developed countries and indicate a high completeness.10 If cases with epilepsy are missed in the Danish National Hospital Register, the relative risk of epilepsy would be aff ected only if the incomplete capture of patients diff ers between those with and without brain injury. Patients with head injury may be followed more closely than the general population, which might increase the completeness and overestimate the relative risk of epilepsy after brain injury. However, the eff ect of this bias is likely to decrease over time.

Although, most patients with epilepsy are cared for on an outpatient basis, the incidence estimate only increased by 17% after inclusion of outpatients.10 Hence, most outpatients with epilepsy are also admitted to hospital for that or other reasons and, thereby, included in the National Hospital Register. Some patients with severe brain injury live in care homes in the community after their condition has stabilised, but these patients have the same access to the hospital system as patients without brain injury, and thus we think that they do not have a decreased likelihood of being registered with epilepsy in the Danish National Hospital Register.

People were censored when they died or left Denmark permanently, but less than 3% of the entire cohort did so.8 Some people may have had a brain injury or epilepsy during a short stay abroad; but numbers are likely to be very low, and most of these will be treated in Danish hospitals or outpatient clinics when they return to Denmark. Bias due to selection of study participants is therefore an unlikely explanation for our fi ndings. In comparison, 1139 (25%) patients of a total population of 4541 in the Rochester study were lost to follow-up due to migration from Minnesota.1

A previous study assessed the validity of the hospital codes for brain injury (ICD-8: 851–854) showing that the diagnoses were confi rmed in about 88% of cases.29 However, clinical discrimination between diff erent types of brain injury is diffi cult and the defi nitions vary between countries.14 Brain injuries that at fi rst seem mild can turn out to be severe. In a study of 24 patients with post-traumatic amnesia lasting more than 1 week, four had initially been diagnosed with skull fracture and four with concussion.14 Although, there is debate about the importance of post-traumatic amnesia in the diagnosis of patients with mild brain injury,14,30 some patients diagnosed with mild head injury might actually suff er from more severe brain injury, which would likely lead to an overestimated risk of epilepsy associated with mild brain injury.

Articles


In the study period (1977–2002), the incidence of mild brain injury decreased,31 probably because fewer children were injured or because the need for observation decreased after introduction of new diagnostic methods, most notably CT and MRI.14,32 If only the more severe mild head injuries were treated in hospitals in later year, it might explain the increased risk of epilepsy after brain injury in 1994–2002.

However, the diagnostic criteria15,29 might have changed when the classifi cation codes were changed from ICD-8 to ICD-10 in 1994,10,11 and completeness of epilepsy in the Danish National Hospital Register might have increased when outpatients were included in 1995.10 In the analyses, we tried to take these factors into account by adjusting for calendar year.

We know, that about 40–50% of all hospitalisations with traumatic brain injuries are related to road-traffi c accidents, 20–25% to falls, 8–10% to fi rearms and assaults, and the remaining related to other causes, such as sporting injuries depending on age and social background.14,33 In this study, we did not have information on cause of brain injury, but prevention measures such as the use of bicycle helmets34,35 might prevent brain injury and subsequent epilepsy, although the eff ectiveness of such measures has been questioned.36,37

Traumatic brain injury is a signifi cant risk indicator for epilepsy many years after the injury. Drug treatment after brain injury with the aim of preventing post-traumatic epilepsy has been discouraging, but our data suggest a long time interval for potential, preventive treatment of high risk patients.

ContributorsJC and MV initiated the study and obtained funding. MV, CBP, MGP, JO,

PSI, and JC designed the study. MGP and CBP constructed the population.

JC, MGP, CBP, and MV analysed the data. JC, MV, and CBP wrote the fi rst

draft; JC wrote the revised versions. All authors interpreted the results,

revised the paper, and approved the fi nal version.

Confl ict of interest statementWe declare that we have no confl ict of interest.

References1 Annegers JF, Hauser WA, Coan SP, Rocca WA. A population-based

study of seizures after traumatic brain injuries. N Engl J Med 1998; 338: 20–24.

2 Frey LC. Epidemiology of posttraumatic epilepsy: a critical review. Epilepsia 2003; 44 (suppl 10): 11–17.

3 Angeleri F, Majkowski J, Cacchio G, et al. Posttraumatic epilepsy risk factors: one-year prospective study after head injury. Epilepsia 1999; 40: 1222–30.

4 Englander J, Bushnik T, Duong TT, et al. Analyzing risk factors for late posttraumatic seizures: a prospective, multicenter investigation. Arch Phys Med Rehabil 2003; 84: 365–73.

5 D’Ambrosio R, Perucca E. Epilepsy after head injury. Curr Opin Neurol 2004; 17: 731–35.

6 Agrawal A, Timothy J, Pandit L, Manju M. Post-traumatic epilepsy: an overview. Clin Neurol Neurosurg 2006; 108: 433–39.

7 Temkin NR. Antiepileptogenesis and seizure prevention trials with antiepileptic drugs: meta-analysis of controlled trials. Epilepsia 2001; 42: 515–24.

8 Pedersen CB, Gotzsche H, Moller JO, Mortensen PB. The Danish Civil Registration System: a cohort of eight million persons. Dan Med Bull 2006; 53: 441–49.

9 Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen JH. The Danish National Hospital Register: a valuable source of data for modern health sciences. Dan Med Bull 1999; 46: 263–68.

10 Christensen J, Vestergaard M, Pedersen MG, Pedersen CB, Olsen J, Sidenius P. Incidence and prevalence of epilepsy in Denmark. Epilepsy Res 2007; 76: 60–65.

11 Christensen J, Vestergaard M, Olsen J, Sidenius P. Validation of epilepsy diagnoses in the Danish National Hospital Register. Epilepsy Res 2007; 75: 162–70.

12 Vestergaard M, Pedersen CB, Sidenius P, Olsen J, Christensen J. The long-term risk of epilepsy after febrile seizures in susceptible subgroups. Am J Epidemiol 2007; 165: 911–18.

13 Sun Y, Vestergaard M, Pedersen CB, Christensen J, Olsen J. Apgar scores and long-term risk of epilepsy. Epidemiology 2006; 17: 296–301.

14 Engberg A. Severe traumatic brain injury—epidemiology, external causes, prevention, and rehabilitation of mental and physical sequelae. Acta Neurol Scand 1995; 164 (suppl): 1–151.

15 Engberg A, Teasdale TW. Traumatic brain injury in children in Denmark: a national 15-year study. Eur J Epidemiol 1998; 14: 165–73.

16 American Congress of Rehabilitation Medicine. Defi nition of mild traumatic brain injury. J Head Trauma Rehabil 1993; 8: 86–88.

17 Pinner M, Børgesen SE, Jensen R, Birket-Smith M, Gade A, Riis JØ. Konsensusrapport om commotio cerebri (hjernerystelse) og det postcommotionelle syndrom. http://www.vfhj.dk/admin/write/fi les/456.pdf (accessed Jan 8, 2008).

18 Breslow NE, Day NE. Statistical methods in cancer research: volume II—the design and analysis of cohort studies. IARC Sci Publ 1987; 82: 1–406.

19 Clayton D, Hills M. Statistical models in epidemiology. Oxford: Oxford University Press, 1993.

20 Breslow NE. Generalized linear models: checking assumptions and strengthening conclusions. Statistica Applicata 1996; 8: 23–41.

21 Pitkanen A, McIntosh TK. Animal models of post-traumatic epilepsy. J Neurotrauma 2006; 23: 241–61.

22 Benardo LS. Prevention of epilepsy after head trauma: do we need new drugs or a new approach? Epilepsia 2003; 44 (suppl 10): 27–33.

23 Behandling af traumatiske hjerneskader og tilgrænsende lidelser. The Danish Board of National Health. 1-208. The Danish National Board of Health, 1997.

24 Kjeldsen MJ, Corey LA, Christensen K, Friis ML. Epileptic seizures and syndromes in twins: the importance of genetic factors. Epilepsy Res 2003; 55: 137–46.

25 Rothman KJ. Modern epidemiology. Boston: Little Brown, 1986.

26 Berkovic SF, Mulley JC, Scheff er IE, Petrou S. Human epilepsies: interaction of genetic and acquired factors. Trends Neurosci 2006; 29: 391–97.

27 Christensen J, Kjeldsen MJ, Andersen H, Friis ML, Sidenius P. Gender diff erences in epilepsy. Epilepsia 2005; 46: 956–60.

28 Commission on Classifi cation and Terminology of the International League Against Epilepsy. Proposal for revised classifi cation of epilepsies and epileptic syndromes. Epilepsia 1989; 30: 389–99.

29 Engberg AW, Teasdale TW. Traumatic brain injury in Denmark 1979–1996: a national study of incidence and mortality. Eur J Epidemiol 2001; 17: 437–42.

30 Bruns TJ, Hauser WA. The epidemiology of traumatic brain injury: a review. Epilepsia 2003; 44: 2–10.

31 Engberg AW, Teasdale TW. Epidemiology and treatment of head injuries in Denmark 1994–2002, illustrated with hospital statistics. Ugeskr Laeger 2007; 169: 199–203.

32 Metting Z, Rodiger LA, De Keyser J, van der Naalt J. Structural and functional neuroimaging in mild-to-moderate head injury. Lancet Neurol 2007; 6: 699–710.

33 Thurman DJ, Alverson C, Dunn KA, Guerrero J, Sniezek JE. Traumatic brain injury in the United States: a public health perspective. J Head Trauma Rehabil 1999; 14: 602–15.

34 Macpherson A, Spinks A. Bicycle helmet legislation for the uptake of helmet use and prevention of head injuries. Cochrane Database Syst Rev 2008; 3: CD005401.

35 Thompson DC, Rivara FP, Thompson R. Helmets for preventing head and facial injuries in bicyclists. Cochrane Database Syst Rev 2000; 2: CD001855.

36 Hewson PJ. Cycle helmets and road casualties in the UK. Traffi c Inj Prev 2005; 6: 127–34.

37 Robinson DL. No clear evidence from countries that have enforced the wearing of helmets. BMJ 2006; 332: 722–25.

Comment


be able to provide a defi nitive treatment decision.9 To refi ne the indication for adjuvant treatment remains the big task for futures studies.

Peter HohenbergerDivision of Surgical Oncology and Thoracic Surgery, Department of Surgery, Medical Faculty Mannheim, University of Heidelberg, D-68135 Mannheim, Germany [email protected]

I have received research grants and honoraria from Novartis.

1 Casali PG, Jost L, Reichardt P, Schlemmer M, Blay J-Y, on behalf of the ESMO Guidelines Working Group. Gastrointestinal stromal tumors: ESMO clinical recommendations for diagnosis, treatment and follow-up. Ann Oncol 2008; 19 (suppl 2): ii35–38.

2 DeMatteo RP, Ballman KV, Antonescu CR, on behalf of the American College of Surgeons Oncology Group (ACOSOG) Intergroup Adjuvant GIST Study Team. Adjuvant imatinib mesylate after resection of localised, primary gastrointestinal stromal tumour: a randomised, double-blind, placebo-controlled trial. Lancet 2009; published online March 19. DOI:10.1016/S0140-6736(09)60500-6.

3 Verweij J, Casali PG, Zalcberg J, et al. Progression-free survival in gastrointestinal stromal tumours with high-dose imatinib: randomised trial. Lancet 2004; 364: 1127–34.

4 Debiec-Rychter M, Sciot R, Le Cesne A, et al, on behalf of the EORTC Soft Tissue and Bone Sarcoma Group, The Italian Sarcoma Group and the Australasian GastroIntestinal Trials Group. KIT mutations and dose selection for imatinib in patients with advanced gastrointestinal stromal tumours. Eur J Cancer 2006; 42: 1093–103.

5 Miettinen M, Lasota J. Gastrointestinal stromal tumors: pathology and prognosis at diff erent sites. Semin Diagn Pathol 2006; 23: 70–83.

6 Corless CL, Schroeder A, Griffi th D, et al. PDGFRA mutations in gastrointestinal stromal tumors: frequency, spectrum and in vitro sensitivity to imatinib. J Clin Oncol 2005; 23: 5357–64.

7 Mussi C, Schildhaus HU, Gronchi A, Wardelmann E, Hohenberger P. Therapeutic consequences from molecular biology for GIST patients aff ected by neurofi bromatosis type 1. Clin Cancer Res 2008; 14: 4550–55.

8 Fletcher CD, Berman JJ, Corless C, et al. Diagnosis of gastrointestinal stromal tumors: a consensus approach. Hum Pathol 2002; 33: 459–65.

9 Gronchi A, Judson I, Nishida T, et al. Adjuvant treatment of GIST with imatinib: solid ground or still quicksand? A comment on behalf of the EORTC Soft Tissue and Bone Sarcoma Group, the Italian Sarcoma Group, the NCRI Sarcoma Clinical Studies Group (UK), the Japanese Study Group on GIST, the French Sarcoma Group and the Spanish Sarcoma Group (GEIS). Eur J Cancer 2009; published online March 16. DOI:10.1016/j.ejca.2009.02.009.

Risk of epilepsy after head traumaHead trauma is an important cause of epilepsy, and knowledge of the extent of the risk of epilepsy after head trauma and the factors that infl uence this risk are essential. In The Lancet today, Jakob Christensen and colleagues1 present their population-based cohort study of more than 1·5 million people born in Denmark between 1977 and 2002, and followed up for that period. 78 572 of them had at least one head injury and 17 470 were diagnosed with epilepsy, of whom 1017 had had a head injury before diagnosis. These researchers obtained the data from the Danish National Hospital Register, which provided diagnostic coding on inpatients from 1977, and outpatients from 1995, on epilepsy and head injury. Family history was ascertained by linkage of data from fi rst-degree relatives. The researchers compared the relative risks of development of epilepsy for people with mild and severe head injury (with or without a family history of epilepsy) on a yearly basis with those for people without head injury, while controlling for age, sex, and calendar year.

Overall, the relative risks of epilepsy were raised about two-fold (relative risk 2·2) after a mild head injury and seven-fold (7·4) after a severe head injury, were slightly greater in women than in men, and increased with older age at time of injury. The rate of development of epilepsy was greatest in the few years after the head injury; for instance, with a greater than fi ve-fold increase

for 2–3 years after a severe head injury, but the excess risk continued for 10 years after mild and severe brain injury—longer than in other studies.2 The incidence of epilepsy was greater in head-injured people with a family history of epilepsy than in those without a family history, with about a six-fold increase in the relative risk of epilepsy after a mild head injury and a ten-fold increase after a severe injury. This fi nding emphasises that the cause of epilepsy is often multifactorial.

Previous studies in this area have been either too small or open to too many methodological criticisms to be deemed to provide defi nitive data. Christensen and co-workers’ investigation is of commendable size and completeness, with an advanced statistical design—as such, it should be accepted as the reference study in the fi eld. This is not to say that there are not methodological criticisms. There are issues inherent in the study design: the diagnosis of epilepsy and the classifi cation of severity of trauma are based on registry data, with all the inaccuracy that this implies; no attempt is made to distinguish between immediate, early, and late epilepsy although these categories have important clinical implications; previously identifi ed risk factors for post-traumatic epilepsy, such as the presence of dural tear, intracranial haemorrhage, and early seizures (<1 week) were not investigated; and no data are provided about the type or severity of the epilepsy.

Published OnlineFebruary 23, 2009

DOI:10.1016/S0140-6736(09)60215-4

See Articles page 1105

Comment


The decision about whether or not to give anti-epileptic drugs prophylactically in patients with head injury is a common clinical dilemma. Christensen and co-workers’ study does not address the value of treatment, but the risk estimates will help patients and doctors make decisions more clearly. The study will also be of value in helping to determine epilepsy risks for medicolegal purposes, by providing a sound basis for determination of cause and compensation and, as such, is a service to social justice. Scientifi c value exists too in the fi nding that the risk of epilepsy was increased for at least 10 years after head injury. Post-traumatic epileptogenesis is thus a long process, which raises

the possibility that neuroprotective measures3 could interfere with this process and thus reduce the risk of epilepsy. Past attempts to prevent epilepsy have been disappointing,4 but these new data suggest that such eff orts should be renewed, to focus particularly on high-risk groups (those with severe head injury, within 2 years of injury, and a positive family history).

Finally, we should note the value of such large-scale epidemiological studies that use pre-existing databases. Such studies are increasingly diffi cult to do in the UK, for example, because of sometimes over-zealous inter-pretation of confi dentiality and consent regulations, and the timidity of the bureaucratic processes. In the UK, we have reached a situation in which, in large swathes of clinical epidemiological research, the baby is being well and truly thrown out with the bathwater, to the detriment of patients and the acquisition of benefi cial knowledge.5

*Simon Shorvon, Aidan NeliganUniversity College London Institute of Neurology, National Hospital for Neurology and Neurosurgery, London WC1N 3BG, [email protected]

We declare that we have no confl ict of interest.

1 Christensen JC, Pedersen MG, Pedersen CB, Sidenius P, Olsen J, Vestergaard M. Long-term risk of epilepsy after traumatic brain injury in children and young adults: a population-based cohort study. Lancet 2009; published online Feb 23. DOI:10.1016/S0140-6736(09)60214-2.

2 Annegers JF, Hauser WA, Coan SP, Rocca WA. A population-based study of seizures after traumatic brain injuries. N Engl J Med 1998; 338: 20–24.

3 Temkin NR. Antiepileptogenesis and seizure prevention trials with antiepileptic drugs: meta-analysis of controlled trials. Epilepsia 2001; 42: 515–24.

4 Temkin NR, Dikmen SS, Wilensky AJ, Keihm J, Chabal S, Winn HR. A randomized, double-blind study of phenytoin for the prevention of post-traumatic seizures. N Engl J Med 1990; 323: 497–502.

5 Metcalfe C, Martin RM, Noble S, et al. Low risk research using routinely collected identifi able health information without informed consent: encounters with the Patient Information Advisory Group. J Med Ethics 2008; 34: 37–40.

Subdural haematoma (red) in 10-year-old boy following trauma

Scie

nce

Phot

o Li

brar

y

Elimination of blinding trachoma revolves around childrenBlinding trachoma is a terrible disease. The intense conjunctival infl ammation in young children causes conjunctival scarring, leading in adult life to inturned eyelashes (trichiasis) that rub on the eye and cause painful blindness. In The Lancet today, Jenafi r House and colleagues report a study in hyperendemic communities in Ethiopia.1 In such areas more than half of children are aff ected, almost every adult has scarring, and 10–20% of

older people have trichiasis. Trachoma is now restricted to poor developing areas, having disappeared from Europe and North America where only a century ago it was a major problem.

Chlamydia trachomatis, the causative bacterium for trachoma, has evolved with human beings and their vertebrate ancestors since Jurassic times.2 It has developed an eff ective host–parasite relation over a

See Articles page 1111

B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

ava i l ab l e a t www.sc i enced i rec t . com

www.e l sev i e r. com/ l oca te /b ra in res

Research Report

Neuroprotective effect of memantine combined withtopiramate in hypoxic–ischemic brain injury

Chunhua Liu, Niyang Lin⁎, Beiyan Wu, Ye QiuDepartment of Pediatrics, The First Affiliated Hospital of Shantou University Medical College, 515000, Shantou, China

A R T I C L E I N F O

⁎ Corresponding author. Fax: +86 754 8898034E-mail address: [email protected] (

0006-8993/$ – see front matter © 2009 Elsevidoi:10.1016/j.brainres.2009.05.071

A B S T R A C T

Article history:Accepted 20 May 2009Available online 6 June 2009

Glutamate receptor-mediated neurotoxicity is a major mechanism contributing to hypoxic–ischemic brain injury (HIBI). Memantine is a safe non-competitive NMDA receptor blockercharacterized by its low affinity and fast unblocking kinetics. Topiramate is an AMPA/KAreceptor blocker and use-dependent sodium channel blocker with several other neuropro-tective actions and little neurotoxicity. We hypothesized that the coadministration ofmemantine and topiramate would be highly effective to attenuate HIBI in neonatal rats.Seven-day-old Sprague–Dawley rat pups were subjected to right common carotid arteryligation and hypoxia for 2 h, and then were randomly and blindly assigned to one of fourgroups: vehicle, memantine, topiramate and combination group. Brain injury was evaluatedby gross damage and weight deficit of the right hemisphere at 22d after hypoxic-ischemia(HI) and byneurofunctional assessment (foot-fault test) at 21dpost-HI. Acute neuronal injurywas also evaluated by microscopic damage grading at 72 h post-HI. Results showed thecombination of memantine and topiramate improved both pathological outcome andperformance significantly. The drug-induced apoptotic neurodegeneration was assessed byTUNEL staining at 48 h post-HI and the result showed no elevated apoptosis in all observedareas. The result of the experiment indicates the combination therapy is safe and highlyeffective to reduce brain damage after HIBI.

© 2009 Elsevier B.V. All rights reserved.

Keywords:NeuroprotectionMemantineTopiramatePharmacologyCombination therapyGlutamate receptorApoptosis

1. Introduction

The excessive glutamate release and overactivation of gluta-mate receptors are crucial contributors to the pathogenesis ofHIBI. They cause amassive influx of sodium (Na+) and calcium(Ca2+) that triggers a cascade of biochemical events, and leadto neuronal necrosis and apoptosis in many types of cells inneonatal brain. There are three subtypes of ionotropicglutamate receptors involved, namely N-methyl-D-aspartate(NMDA) receptors, α-3-amino-hydroxy-5-methyl-4-isoxazolepro-pionic acid (AMPA) receptors and kainate (KA) receptors.Excitotoxic injury occurs secondary to glutamate-triggeredCa2+ influx through any of three routes: NMDA channels,

7.N. Lin).

er B.V. All rights reserved

voltage-sensitive Ca2+ channels, and Ca2+-permeable AMPA/KA channels (Lu et al., 1996). The excitotoxic overactivation ofNMDA and AMPA/KA glutamate receptors provokes furtherglutamate release and further NMDA and AMPA/KA receptorstimulation, and it forms a positive feedback cycle making thecondition worse (Villmann and Becker, 2007). To break thisvicious cycle, researchersutilizedmanyNMDAand/orAMPA/KAreceptor antagonists, but most of them have severe side effects(Haberny et al., 2002; Puka-Sundvall et al., 2000). For example, awidely used NMDA antagonist MK-801 was found inducingwidespread apoptotic neurodegeneration and impairing manynormal neuronal functions in developing rat brain (Ikonomidouet al., 1999). As NMDA receptors are essential for normal

.

mailto:[email protected]

http://dx.doi.org/10.1016/j.brainres.2009.05.071

Table 1 – Neurologic damage score.

Group No. Normal=1 Mild=2 Moderate=3 Severe=4 p ⁎

Vehicle 19 4 4 3 8 NSMemantine 24 10 7 4 3 <0.05Topiramate 21 5 5 5 6 >0.05Combination 24 13 6 4 1 <0.01

The number of pups receiving the designated gross damage score by a blinded observer.⁎ p value, memantine, topiramate, combination vs. vehicle.

Fig. 1 – The percentage of reduction in right cerebralhemisphere weight measured using the left hemisphereweight as standard. The animal numbers are as described inthe result. The percentage of reduction in right hemisphereweight was significantly decreased in the combination groupcompared with the vehicle group (**p<0.01 vs. vehicle). Thepercentage of reduction in right hemisphere weight wassignificantly decreased in the memantine group comparedwith the vehicle group (*p<0.05 vs. vehicle). Data arepresented as mean±S.D.

174 B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

physiological processes in brain development, including theproliferation,migration, survival and differentiation of neurons,blockade of excessive NMDA receptor activity must be achievedwithout affecting normal brain functioning (Kohr, 2007).

Recently, increasing evidence based on molecular studiessuggests that memantine, an uncompetitive NMDA receptorblocker with fast channel unblocking kinetics to prevent it fromoccupying the channels and interfering with normal synaptictransmission, is a potent neuroprotectant without above-mentioned side effects (Chen et al., 1992, 1998; Chen and Lipton,2005; Johnson and Kotermanski, 2006). In contrast to MK-801and ketamine, memantine shows unusual clinical tolerance inthe treatment of moderate-to-severe Alzheimer's disease inadults through its low affinity and relatively fast unblockingkinetics (de Lima et al., 2000; Lipton, 2004; Lipton, 2006). As aneuroprotective agent, memantine can reduce functional aswell asmorphological sequelae induced by ischemia (Block andSchwarz, 1996; Chen et al., 1998). A recent study showed theNMDA receptor blockade with memantine could provide aneffective pharmacological prevention of periventricular leuko-malacia (PVL) in the premature infant (Manning et al., 2008).

Topiramate, a well tolerated antiepileptic drug (AED) usedclinically, confers neuroprotection by blocking AMPA/KAreceptors and use-dependent Na+ channel in developing ratbrain without serious side effects compared to conventionalanticonvulsants (Noh et al., 2006). Topiramate has anti-excitotoxic properties, because it protects against motorneuron degeneration. The other neuroprotective effects oftopiramate include positive modulation of gamma-aminobu-tyric acid (GABA) receptors, increase of seizure threshold andso on (Pappalardo et al., 2004). Furthermore, Topiramate alsoprotects preoligodendrocytes against excitotoxic cellulardeath in whitematter lesions and prevents the periventricularwhite matter from the damage induced by an AMPA/KAagonist in newbornmice (Follett et al., 2004; Sfaello et al., 2005).

Due to the complex pathological mechanisms in HIBIdescribed above, combination therapy or multimodal target-ing is thought to be a key future approach to provide effectiveneuroprotection. Most promising combination should targetdifferent neuroprotective mechanisms, expand the therapeu-tic time window, and alleviate the possibility of side effects(Rogalewski et al., 2006). Studies on the mechanisms of thesuperfamily of glutamate receptors revealed that NMDA andAMPA glutamate receptors showed a fine-tuned interaction atthe glutamatergic synapse: the rapid activation and brief opentime of AMPA receptors facilitates unblock of NMDA receptors(Villmann and Becker, 2007). Functional interdependence ofAMPA and NMDA receptors has been proven by experimentswhere a transient synaptic activation of NMDA receptorsreliably induces a long-term potentiation phenomenon,

associated with an increase in the intensity and number ofsynaptic AMPA-receptor clusters (Liao et al., 2001; Liu et al.,2004a). These findings suggest that it will be more effectiveand beneficial to block both NMDA and AMPA/KA receptors bycombination of different glutamate receptor antagonists.

Based on the pharmacology and mechanism studies, wedesigned the experiments to evaluate the efficacy of thecombination therapy by measuring gross brain damage, brainweight deficit in the right hemisphere and regional neuronalinjury. Besides themorphologic and histopathologic measure-ment, a neurofunctional test was performed to verify theresults. To ensure therapeutic safety, the possible drug-induced apoptosis was assessed even though the two drugswere approved safe and efficient in their respective therapeu-tic categories (Chen et al., 1998; Glier et al., 2004).

2. Results

2.1. Gross brain damage grading

The neurologic damage score was determined by an observerblind to the drug treatment of the rat pups. Table 1 shows theneurologic damage scores in each group. The neurologicdamage score was significantly higher in the vehicle-treatedgroup (2.79±1.23, n=19) than that in the combination-treated

Fig. 2 – Microscopic brain damage scores in the cortex, hippocampus, striatum, thalamus. Data are presented as mean±S.D.*p<0.05 vs. vehicle, **p<0.01 vs. vehicle.

Fig. 3 –Number of foot-faults in each group. The combinationgroup had significantly fewer foot-faults than the vehiclegroup. Data are presented as mean±S.D. *p<0.05 vs. vehicle.

175B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

group (1.71±0.91, n=24, p<0.01 versus vehicle). The neurologicdamage score was significantly higher in the vehicle-treatedgroup than that in the memantine-treated group (2.00±1.06,n=24, p<0.05 versus vehicle). The neurologic damage score inthe topiramate-treated group (2.57±1.17, n=21, p>0.05 versusvehicle) was lower but not statistically significant comparedwith the vehicle-treated group.

2.2. Brain weight deficit

Fig. 1 shows the weight deficit in the right hemisphere relativeto the left hemisphere. The weight deficit in the combination-treated group (9.2±2.5%, n=24, p<0.01 versus vehicle) wassignificantly reduced compared with the vehicle-treatedgroup (26.9±4.1%, n=19). The weight deficit in the meman-tine-treated group (16.3±3.2%, n=24, p<0.05 versus vehicle)was significantly reduced compared with the vehicle-treatedgroup. The weight deficit in the topiramate-treated group(21.5±4.0%, n=21, p>0.05 versus vehicle) was reduced butnot statistically significant compared with the vehicle-treated group. Body weights of rat pups in each group wererecorded and analyzed. Results showed that the bodyweights of the treated groups were not significantly differentfrom the vehicle-treated group at 1, 3, 7, 14 and 22 days afterinjury (data not shown). Mortality rates were not signifi-cantly different in four groups, although there was a trendtoward reduced mortality in the combination group.

2.3. Microscopic brain damage grading

Themicroscopicbraindamagescore (histopathologic score)wasdetermined by anobserver blind to thedrug treatment of the ratpups. Fig. 2 shows the microscopic brain damage score in eachgroup. The histopathologic score in the memantine-treated

group (2.15±0.52 and 1.51±0.47, n=12, p<0.05 and p<0.05 versusvehicle) was significantly lower compared with the vehicle-treated group (4.15±0.73 and 3.38±0.72, n=10) in the cortex andthalamus.Thehistopathologic score in thecombination-treatedgroup (1.91±0.51, 1.45±0.49 and 0.91±0.42, n=12, p<0.05,p<0.05 and p<0.01 versus vehicle) was significantly lowercompared with the vehicle-treated group (4.15±0.73, 3.68±0.62and 3.38±0.72, n=10) in the cortex, hippocampus and thalamus.In the striatum, the histopathologic score in the combination-treated group was lower but not statistically significantcompared with the vehicle-treated group.

2.4. Foot-fault test

Fig. 3 shows the number of foot-faults in each group. Thenumber of foot-faults per pup was significantly greater in thevehicle-treated group (8.62±1.51, n=10) than that in the

Fig. 4 – The numbers of TUNEL-positive apoptotic cells in the cortex, the CA1, CA3 and dentate gyrus of the hippocampus, thestriatum and the subcortical whitematter in the vehicle, memantine, topiramate and combination group. Data are presented asmean±S.D. *p<0.05 vs. vehicle, **p<0.01 vs. vehicle.

176 B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

combination-treated group (4.26±0.93, n=12, p<0.05 versusvehicle). The number of foot-faults per pup was significantlygreater in the vehicle-treated group than that in the meman-tine-treated group (4.66±1.03, n=12, p<0.05 versus vehicle).The number of foot-faults per pupwas less but not statisticallysignificant in the topiramate-treated group (6.94±1.22, n=11)compared with in the vehicle-treated group.

2.5. TUNEL-positive cell counting

The numbers of TUNEL-positive apoptotic cells of each groupare presented in Fig. 4 and areas examined for drug-induced

Fig. 5 – Areas of the brain examined for neuronal injury anddrug-induced apoptosis. CX=cortex CA1=hippocampus CA1CA3=hippocampus CA3 Den=dentate gyrus ST=striatumTH=thalamus.

apoptosis are shown in Fig. 5. In all observed areas, thenumbers of apoptotic cells in the treated group (single orcombined) were not significantly increased comparedwith thevehicle-treated group. In the CA1 sector of the hippocampus,The numbers of apoptotic cells in the combination-treatedgroup (31.2±20.7 and 45.5±31.2, n=12, p<0.01 and p<0.01versus vehicle) were significantly reduced compared with thevehicle-treated group (82.1±32.6 and 175±48.2, n=12). In theCA1 sector of the hippocampus and the subcortical whitematter, The numbers of apoptotic cells in the memantine-treated group (50.5±28.3 and 99.8±38.7, n=12, p<0.05 andp<0.05 versus vehicle) were significantly reduced comparedwith the vehicle-treated group. In other areas, no significantdifferences were found between any of the treated groups(single or combined) and the vehicle group. Fig. 6 shows somesample pictures of apoptotic cells in the CA1 sector of thehippocampus.

3. Discussion

The present study shows for the first time to our knowledgethat the combination of memantine and topiramate exertsenhanced protection of neurons against HIBI in vivo, comparedwith each of these agents alone. In this study, we measuredbrain damage in each group by using the gross anatomicmethod of Palmer et al. at 22d post-HI. By delaying assessmentuntil 22d after HI, we included very late cell death that reflectsoverall neuroprotective effect of the drugs in a relatively longperiod. We also examined the brain weight deficit presentedby the loss of brain weight on the ipsilateral side relative to thecontralateral side. Results showed the combination therapysignificantly reduced the degree of brain injury in this model.Besides the morphologic examinations, we applied the foot-

Fig. 6 – Sample pictures of TUNEL-positive apoptotic cells in the CA1 sector of the hippocampus in the (A) vehicle,(B) memantine, (C) topiramate and (D) combination group. Original magnification, ×400.

177B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

fault test to evaluate sensorimotor function of the rat pups at21d post-HI. Foot-faults per pup in the combination groupwere significantly less than that in the vehicle group. Thefunctional outcome was consistent with the morphologicfindings in the long-term perspective. The short-term effect ofthe combination therapy was evaluated by microscopic braindamage scoring at 72 h post-HI. Results showed that thecombination therapy reduced neuronal injury significantly inthe cortex, hippocampus and thalamus.

Neuronal cell death after HI has generally been attributedto either rapid necrosis or delayed apoptosis. There is no doubtthat necrosis plays major role in the course. But the develop-ing brainmay have good plasticity and a high capacity for self-repair (Daval et al., 2004; Grafe, 1994). After most compensa-tory and reparative phrases have passed, there are at leastthree different end points should be taken into account inassessment: the long-term deficit of brain tissue, the func-tional consequences of the brain injury and the acute extent ofbrain injury (Bona et al., 1997). The quantitive assessment ofbrain weight deficit and gross brain damage used in this studycan accurately evaluate neuroprotective effects of glutamateantagonists against NMDA-mediated brain injury in vivo(Andine et al., 1990; McDonald et al., 1989a). On the otherhand, behavioral consequences after HIBI are essential toreveal the true functional disability and to study the effects ofdrug intervention. In this study, the foot-fault test was done at21d post-HI to evaluate the long-term functional outcome.Different from other cognitive function tests (Morris watermaze, etc) related mostly to the hippocampus formation, thefoot-fault test correlates with brain lesion in the cerebralcortex which is the most constantly affected region in bothmild and severe HIBI in this model (Bona et al., 1997). Short-term effect of the therapy was evaluated by a scoring systemon neuronal injury in 4 main regions of the rat brain at 72 h

post-HI. Because short-term neuronal injury in the developingbrain after HI is caused by both early and delayed neurode-generation, the onset of damage in different regions of thebrain is time-dependent and progressive, and it has an unevendistributionwithin regions (Northington et al., 2001). However,72h (3d) post-HI seems an appropriate time point to evaluateshort-term neuronal injury after insult in this model (Feng etal., 2005, 2008; Manning et al., 2008; Zhu et al., 2004).

In our experiment, the time window and doses ofmemantine and topiramate were chosen according to ageneral purpose to achieve an application for potential clinicaluse. Based on published data of rat pharmacokinetics anddose–response studies, 20 mg/kg dose of memantine canprovide minimal neuroprotection (Chen et al., 1998; Hesselinket al., 1999). Considering the short therapeutic time window(Culmsee et al., 2004) and the confirmed neuroprotectiveeffects of memantine at 20 mg/kg dose in HI and PVL model,we administered the 20 mg/kg loading dose of memantineimmediately after HI in the treatment. Topiramate (loadingdose 50 mg/kg, maintenance dose 20 mg/kg/day) can reduceneuronal cell loss significantly but increase apoptosis in thefrontal white matter in newborn piglets (Schubert et al., 2005).Furthermore, topiramatemay cause neurodegeneration in thedeveloping rat brain only at doses at and above 50 mg/kg(Glier et al., 2004). The reason why topiramate at doses above50 mg/kg can protect neurons but increase apoptosis mayrelate to two mechanisms. The first one is the blockade ofAMPA/KA receptors lack of interference with NMDA-receptorsignaling (Gibbs et al., 2000). Topiramate cannot provideneuroprotection only through AMPA/KA receptor channelunless it reaches threshold dosage. The second one is thedepression of the endogenous neurotrophin system in thebrain which may account for the proapoptotic effect (Bittigauet al., 2002). In a gerbil model, topiramate was found reducing

178 B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

hippocampal neuronal damage in dose-dependent manner(Lee et al., 2000). Based on the dose–response studies and ourpreliminary experiment, we chose 40 mg/kg as the loadingdose for topiramate. The dose of topiramate (loading dose40 mg/kg; maintenance dose 20 mg/kg/day) was provenconsiderably safe but unlikely to be neuroprotective.

Although the mechanisms underlying the neuroprotectionare not fully understood, the results demonstrate that asynergistic reduction in brain damage can be achievedeffectively by memantine combined with topiramate. Theneuroprotective actions and unique characteristics of thesetwo drugs may account for the experimental outcome. It iswell documented that memantine antagonizes NMDA recep-tor activation by inhibiting the influx of Ca2+ through thischannel (Johnson and Kotermanski, 2006). As an open-channel blocker, memantine can provide neuroprotectionwithout interference with the normal brain development(Parsons et al., 1999). The favorable kinetics of memantineinteractionwith NMDA channelsmay be partly responsible forits high index of therapeutic safety, and itmakesmemantine acandidate drug for use in many NMDA receptor-mediatedhuman CNS disorders (Johnson and Kotermanski, 2006;Lipton, 2004). In a four-vessel-occlusion (4VO) global ischemicmodel, neuronal damage in the CA1 sector of the hippocam-pus and in the striatum produced by 4VO was significantlyattenuated by 20mg/kgmemantine (Block and Schwarz, 1996).Memantine has been used clinically for excitotoxic disordersat neuroprotective doses administered up to 2 h afterinduction of HI in immature and adult rats. At neuroprotectiveconcentrations, memantine results in few adverse side effectsand displays virtually no effects on Morris water mazeperformance or on neuronal vacuolation (Chen et al., 1998).Rosi et al. found thatmemantine protects against LPS-inducedneuroinflammation, and confers neural and cognitive protec-tion (Rosi et al., 2006). Furthermore, NMDA receptor blockadewith memantine can provide an effective pharmacologicalprevention of PVL in the premature infant without affectingnormal myelination or cortical growth (Manning et al., 2008).

Topiramate is a novel broad spectrum antiepileptic drug(AED) used clinically in adults and children older than 2 years.Amongst new-generation AEDs examined for neurotoxicity inneonatal rats, topiramate holds promise for minimizing therisk of neuronal death without side effects such as theimpairment of cognitive performance (Cha et al., 2002; Glieret al., 2004; Mellon et al., 2007). Pharmacological actions oftopiramate include positive modulation of GABA receptors,inhibition of the AMPA/KA glutamate receptor subtypes andblockade of a use-dependent Na+ channel (Schubert et al.,2005). Noh and his coworkers reported the co-treatment oftopiramate and an NMDA receptor antagonist D-AP5 greatlyincreased the number of viable neurons in oxygen–glucosedeprivated cells. The experiment determined that neuropro-tective effect of topiramate was mainly mediated by theinhibition of AMPA glutamate receptors (Noh et al., 2006).Topiramate blocks the spread of seizures caused by transientglobal cerebral ischemia, and reduces the abnormally highextracellular levels of glutamate in the hippocampus in theimmature rat spontaneous epileptic model by blocking AMPAreceptors (Koh et al., 2004). It also affects the expression ofglutamate transporters (GLAST and GLT-1) which are respon-

sible for the inactivation of glutamate as a neurotransmitter(Poulsen et al., 2006). Moreover, topiramate was foundeffective in attenuating seizure-induced neuronal cell deathand reducing KA-induced Phospho-extracellular signal-regu-lated kinase-immunoreactive (p-Erk IR) in the CA3 region ofthe hippocampus (Park et al., 2008). In a rat pup model of PVL,topiramate has been demonstrated effective to attenuateAMPA/KA receptor-mediated cell death and Ca2+ influx, aswell as KA-evoked currents in developing oligodendrocytes(Follett et al., 2004).

Many studies suggest that combination of drugs mayproduce greater toxicity than individual ones. Thus, the safetyof combination therapies should be most concerned, whenthese animal findings are intended for extrapolating to apediatric surgical patient population (Bittigau et al., 2002). Therat is most sensitive to NMDA receptor-mediated neurotoxi-city during early neuronal pathway development, referred toas the “brain-growth spurt period” or period of synaptogen-esis. (Haberny et al., 2002). Blockade of NMDA receptors up to4 h is sufficient to trigger apoptotic neurodegeneration in thedeveloping brain (Ikonomidou et al., 1999). In consideration ofthe possible neurotoxicity caused by the coadminstration ofdrugs and the complicated interaction between NMDA recep-tor blocker and AMPA receptor blocker, we examined thepossible drug-induced neuronal apoptosis by TUNEL stainingat 48 h post-HI even through the two drugs are proven safe atthe given doses respectively (Chen et al., 1998; Glier et al.,2004). The time course of apoptotic injury varies regionallybecause HI damage generally evolves more rapidly in theimmature brain than its adult counterpart. Injury in the cortexand striatum occurs in a biphasic manner, where the earlyphase (by 3 h) is classified as necrosis and the later phase (by48 h) displays signs of apoptosis (Northington et al., 2001).Nakajima et al. found that the density of caspase-3 immunor-eactivity was enhanced in the frontal, parietal, and cingulatecortex and in the striatum 24 h after hypoxic ischemic injury.In the CA3 sector of the hippocampus, the dentate gyrus,medial habenula and laterodorsal thalamus, the density ofapoptotic cells was highest at 24–72 h after HI and thendeclined. In thalamus, increased caspase-3 immunoreactivitywas distributed in lateral, laterodorsal, and reticular nucleiwith a peak in density at 48 h after HI. In hippocampus,intense caspase-3 immunoreactivity was present in CA1 andin the dentate gyrus at 48 h after insult but had nearlydisappeared by 7d after HI injury (Nakajima et al., 2000).Based on all these results on apoptotic injury, the time point(48 h post-HI) was chosen to examine the apoptoticneurodegeneration.

In this experiment, massive cellular apoptosis was notfound in all observed areas in the treated groups, andapoptosis was reduced in the CA1 sector of the hippocampusand the subcortical white matter in the combination groupcompared with the vehicle group. The safe dosing regimenand anti-apoptotic actions of memantine and topiramatemaycontribute to the results synergistically. Regional patterns ofneuronal death can also be detected by expression of caspase-3, a cysteine protease involved in the execution phase ofapoptosis. Immunocytochemical and Western blot analysesshow increased caspase-3 expression in damaged hemi-spheres 24 h to 7d after HI. Reduced caspase-3 activity has

179B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

been shown to be associated with neuroprotection (Endres etal., 1998; Puka-Sundvall et al., 2000). Memantine (20 mg/kg, i.p.) can prevent isoflurane-induced caspase-3 activation andapoptosis in vivo and in vitro. The results also indicated thatisoflurane-induced caspase activation and apoptosis aredependent on cytosolic calcium levels (Zhang et al., 2008). Inrecent years, many studies focus on the protection of whitematter because the importance of PVL pathophysiology hasbeen realized gradually (Khwaja and Volpe, 2008; Volpe, 2008).NMDA receptor blockade with memantine acts as an effectivepharmacological contributor with little side effects in attenu-ating white matter injury, and the protective dose ofmemantine does not affect normal myelination or corticalgrowth (Manning et al., 2008; Micu et al., 2006). In ourexperiment, the apoptosis in the subcortical white matterwas reduced significantly in the combination group, which isconsistent with the previous findings on caspase-3 activation.

The present study demonstrated that a synergistic reduc-tion in brain damage could be achieved by combination ofneuroprotective agents targeting different mechanisms.Although an evolving body of work has shown that combina-tion therapy holds promise in the treatment of HIBI, there hasbeen relatively little research on the combination therapy oftwo glutamate receptor antagonists. The combination ofNMDA receptor antagonist MK-801 and AMPA receptorantagonist NBQX shows an “overadditive” effect in cell cultureand focal ischemia model in mice (Lippert et al., 1994). On theother hand, several studies onmemantine or topiramate haveshown multidrug strategies are required for optimal thera-peutic outcome. The combination of memantine and clenbu-terol not only reduces the infarct size but also extends thetherapeutic window of clenbuterol up to 2 h after ischemia(Culmsee et al., 2004). The combination of memantine andcelecoxib shows better effects in neuroprotection and anti-inflammation in intracerebral hemorrhage treatment (Sinn etal., 2007). Combined treatment with topiramate and delayedhypothermia improves both performance and pathologicaloutcome in P15 and P35 rats (Liu et al., 2004b).

Although the present study demonstrates the neuroprotec-tive effect of memantine combined with topiramate, furtherstudies are still needed in two aspects. A full dose–responseexperiment was not performed in the present study, so furtherinvestigation is still needed to determine the most optimaldosing regimen of memantine and topiramate. Noh et al.suggested that the pretreatment with topiramate before HIwas more effective than the post-treatment after HI (Noh et al.,2006). The result implies that the pretreatment with topiramatein the combination therapy can be considered in the future.

Collectively, the present study not only shows a promisingtherapy for neuroprotection, but also proposes a new para-digm for multidrug development which is thought to be apromising approach in the treatment of HIBI.

4. Experimental procedures

4.1. Animal procedures

Seven-day-old rat pups of either sex, weighing between 12 gand 16 g, were used in this study. The rat pups were randomly

assigned to one of the following groups: vehicle group (saline),memantine group, topiramate group, combination group(memantine and topiramate). All animal experiments fol-lowed a protocol approved by the ethical committee on animalresearch at our institution. The neonatal HI brain damage wasinduced according to the modified Levine–Rice procedure(Northington, 2006; Rice et al., 1981; Vannucci and Vannucci,2005). For short, rat pups were anaesthetized by halothaneinhalation and duration of anesthesia was less than 5 min.The right common carotid artery was dissected, and doublyligated. One hour later, rats were then placed in a plasticchamber (37 °C) and exposed to 8% oxygen and 92% nitrogenfor 2 h. After this hypoxic exposure, the pups were returned totheir dams for 2 h recovery.

4.2. Drug administration

During recovery from HI, drugs were injected intraperitone-ally: vehicle group received vehicle (0.5 ml 0.9% saline)immediately after HI; memantine group received 20 mg/kgloading dose immediately after HI, then 1 mg/kg maintenancedose at 12 h intervals for 48h; topiramate group received40 mg/kg loading dose then 10 mg/kg maintenance dose onthe same schedule as memantine; combination groupreceived both memantine and topiramate, the drug dosesand schedule were the same as above.

4.3. Gross brain damage grading

To quantify the severity of brain damage, rat pups weredecapitated at 22d after HI and their brains were rapidlydissected and frozen (Uhm et al., 2003). Then brains werescored normal, mild, moderate or severe by a blinded observeraccording to themethod of Palmer et al. (1990). The neurologicdamage scores were given according to the following criteria.Normal (1) is no reduction in the size of the right hemisphere,mild (2) is visible reduction in right hemisphere size, moderate(3) is large reduction in hemisphere size from a visible infarctin the right parietal area and severe (4) is near total destructionof the hemisphere.

To measure the loss of hemispheric weight, the brain wasdivided into two hemispheres andweighed after removing thecerebellum and brainstem. Results are presented as thepercent loss of hemispheric weight of the right side relativeto the left [(left−right) / left×100]. The HI model used in thisstudy results in brain damage only on the ipsilateral side, thusthe loss of hemispheric weight can be used as a measure ofbrain damage in this model (Rice et al., 1981). Because thebrain weighs approximately 1 g/ml, weight loss is equivalentto volume loss. According to the method by McDonald et al.,the loss of brain weight on the ipsilateral side relative to thecontralateral side is highly correlated with cellular damage(McDonald et al., 1989b). For short, weighing can assess thedegree of brain damage.

4.4. Microscopic brain damage grading

Microscopic examination of the tissues was carried out toverify that the gross changes were a reflection of the expectedhistopathologic changes. The rat pups were anesthetized

180 B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

with pentobarbital 3 days after injury. Their brains wereperfusion fixed by cardiac puncture. They were flushed withsaline then fixed with 10% buffered formalin. After removal,the brains were stored in 10% buffered formalin. Sectionswere then embedded with paraffin. Five-micron coronalsections were cut in the parietal region aiming for theequivalent of Bregma −4.3 to −4.5 mm in the adult rat (Krugeret al., 1995) and then stained with hemotoxylin and eosin.Cerebral cortex, hippocampus, striatum, thalamus wasscored from 0 to 5 by an observer blind to the treatmentaccording to the method of Cataltepe et al. (1995), where “0” isnormal, “1” is 1–5% of neurons damaged, “2” is 6 to 25% ofneurons damaged, “3” is 26–50% of neurons damaged, “4” is51–75% of neurons damaged, “5” is >75% of neuronsdamaged.

4.5. Neurofunctional assessment: foot-fault test

The foot-fault test was performed at 21d post-HI according toa published method (Bona et al., 1997). Rats were placed onan elevated stainless steel grid floor 50×40 cm, 1 m abovethe floor with 3 cm2 holes and a wire diameter of 0.4 cm.Each pup was placed on the grid and observed for 2 min. Thefoot-fault was defined as when the animal misplaced a fore-or hindlimb and the paw fell through between the grid bars.The excess of left (contralateral foot-faults) to right (ipsilat-eral foot-faults) was recorded. Only the side difference offoot-faults was used for the statistical evaluation to elim-inate the influence of the extent of activity in different rats(Barth and Stanfield, 1990).

4.6. TUNEL staining and apoptotic cell counting

We applied the Terminal deoxynucleotidyl transferase-mediated dUTP Nick End Labeling (TUNEL) staining todetect drug-induced apoptosis at 48 h after HI. All proce-dures were performed following the manufacturer's instruc-tions (In Situ Cell Apoptosis Detection Kit I, POD; Boster,Wuhan, China).

Cell counting was performed in the cortex, hippocampus,striatum and the subcortical white matter. The hippocampuswas divided into the CA1, CA3 and dentate gyrus subfields.Positive cells were counted at 400× magnification (one visualfield=0.196 mm2). By use of the ImageJ software, all analyseswere done by an individual who was unaware of treatmentconditions. The average number of TUNEL-positive cells wascalculated from at least three sections within each region foreach animal. In the hippocampus subfields, counting wasperformed throughout the entire region. In the cortex,striatum and the subcortical white matter, three visual fieldswere counted as average number per visual field. Only thedensely stained cells were counted as TUNEL-positive,slightly TUNEL-stained cells were not (Zhu et al., 2004).

4.7. Statistical analysis

Data are presented as mean±S.D., if not otherwise indicated.Comparisonswere performed by one-way ANOVAwith Fisher'spost hoc test. Differences were considered significant whenp<0.05.

Acknowledgments

This work was partly supported by Natural Science Founda-tion of Guangdong Province, China. We gratefully thankTianhua, Huang for his technical assistance.

R E F E R E N C E S

Andine, P., Thordstein, M., Kjellmer, I., Nordborg, C., Thiringer, K.,Wennberg, E., Hagberg, H., 1990. Evaluation of brain damage ina ratmodel of neonatal hypoxic-ischemia. J. Neurosci. Methods35, 253–260.

Barth, T.M., Stanfield, B.B., 1990. The recovery of forelimb-placingbehavior in rats with neonatal unilateral cortical damageinvolves the remaining hemisphere. J. Neurosci. 10, 3449–3459.

Bittigau, P., Sifringer, M., Genz, K., Reith, E., Pospischil, D.,Govindarajalu, S., Dzietko, M., Pesditschek, S., Mai, I.,Dikranian, K., Olney, J.W., Ikonomidou, C., 2002. Antiepilepticdrugs and apoptotic neurodegeneration in the developingbrain. Proc. Natl. Acad. Sci. U. S. A. 99, 15089–15094.

Block, F., Schwarz, M., 1996. Memantine reduces functional andmorphological consequences induced by global ischemia inrats. Neurosci. Lett. 208, 41–44.

Bona, E., Johansson, B.B., Hagberg, H., 1997. Sensorimotor functionand neuropathology five to six weeks after hypoxia–ischemiain seven-day-old rats. Pediatr. Res. 42, 678–683.

Cataltepe, O., Vannucci, R.C., Heitjan, D.F., Towfighi, J., 1995. Effectof status epilepticus on hypoxic–ischemic brain damage in theimmature rat. Pediatr. Res. 38, 251–257.

Cha, B.H., Silveira, D.C., Liu, X., Hu, Y., Holmes, G.L., 2002. Effect oftopiramate following recurrent and prolonged seizures duringearly development. Epilepsy Res. 51, 217–232.

Chen, H.S., Lipton, S.A., 2005. Pharmacological implications of twodistinct mechanisms of interaction of memantine withN-methyl-D-aspartate-gated channels. J. Pharmacol. Exp. Ther.314, 961–971.

Chen, H.S., Pellegrini, J.W., Aggarwal, S.K., Lei, S.Z., Warach, S.,Jensen, F.E., Lipton, S.A., 1992. Open-channel block ofN-methyl-D-aspartate (NMDA) responses by memantine:therapeutic advantage against NMDA receptor-mediatedneurotoxicity. J. Neurosci. 12, 4427–4436.

Chen, H.S., Wang, Y.F., Rayudu, P.V., Edgecomb, P., Neill, J.C., Segal,M.M., Lipton, S.A., Jensen, F.E., 1998. Neuroprotectiveconcentrations of the N-methyl-D-aspartate open-channelblocker memantine are effective without cytoplasmicvacuolation following post-ischemic administration and donot block maze learning or long-term potentiation.Neuroscience 86, 1121–1132.

Culmsee, C., Junker, V., Kremers, W., Thal, S., Plesnila, N.,Krieglstein, J., 2004. Combination therapy in ischemic stroke:synergistic neuroprotective effects of memantine andclenbuterol. Stroke 35, 1197–1202.

Daval, J.L., Pourie, G., Grojean, S., Lievre, V., Strazielle, C., Blaise, S.,Vert, P., 2004. Neonatal hypoxia triggers transient apoptosisfollowed by neurogenesis in the rat CA1 hippocampus. Pediatr.Res. 55, 561–567.

de Lima, J., Beggs, S., Howard, R., 2000. Neural toxicity of ketamineand othe NMDA antagonists. Pain 88, 311–312.

Endres, M., Namura, S., Shimizu-Sasamata, M., Waeber, C., Zhang,L., Gomez-Isla, T., Hyman, B.T., Moskowitz, M.A., 1998.Attenuation of delayed neuronal death after mild focalischemia in mice by inhibition of the caspase family. J. Cereb.Blood Flow Metab. 18, 238–247.

Feng, Y., Fratkins, J.D., LeBlanc, M.H., 2005. Estrogen attenuateshypoxic–ischemic brain injury in neonatal rats. Eur. J.Pharmacol. 507, 77–86.

181B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

Feng, Y., Bhatt, A.J., Fratkin, J.D., Rhodes, P.G., 2008.Neuroprotective effects of sodium orthovanadate afterhypoxic–ischemic brain injury in neonatal rats. Brain Res. Bull.76, 102–108.

Follett, P.L., Deng, W., Dai, W., Talos, D.M., Massillon, L.J.,Rosenberg, P.A., Volpe, J.J., Jensen, F.E., 2004. Glutamatereceptor-mediated oligodendrocyte toxicity in periventricularleukomalacia: a protective role for topiramate. J. Neurosci. 24,4412–4420.

Gibbs III, J.W., Sombati, S., DeLorenzo, R.J., Coulter, D.A., 2000.Cellular actions of topiramate: blockade of kainate-evokedinward currents in cultured hippocampal neurons. Epilepsia 41(Suppl 1), S10–S16.

Glier, C., Dzietko, M., Bittigau, P., Jarosz, B., Korobowicz, E.,Ikonomidou, C., 2004. Therapeutic doses of topiramate are nottoxic to the developing rat brain. Exp. Neurol. 187, 403–409.

Grafe, M.R., 1994. Developmental changes in the sensitivity of theneonatal rat brain to hypoxic/ischemic injury. Brain Res. 653,161–166.

Haberny, K.A., Paule, M.G., Scallet, A.C., Sistare, F.D., Lester, D.S.,Hanig, J.P., Slikker Jr, W., 2002. Ontogeny of theN-methyl-D-aspartate (NMDA) receptor system andsusceptibility to neurotoxicity. Toxicol. Sci. 68, 9–17.

Hesselink, M.B., De Boer, B.G., Breimer, D.D., Danysz, W., 1999.Brain penetration and in vivo recovery of NMDA receptorantagonists amantadine and memantine: a quantitativemicrodialysis study. Pharm. Res. 16, 637–642.

Ikonomidou, C., Bosch, F., Miksa, M., Bittigau, P., Vockler, J.,Dikranian, K., Tenkova, T.I., Stefovska, V., Turski, L., Olney,J.W., 1999. Blockade of NMDA receptors and apoptoticneurodegeneration in the developing brain. Science 283,70–74.

Johnson, J.W., Kotermanski, S.E., 2006. Mechanism of action ofmemantine. Curr. Opin. Pharmacol. 6, 61–67.

Khwaja, O., Volpe, J.J., 2008. Pathogenesis of cerebral white matterinjury of prematurity. Arch. Dis. Child., Fetal Neonatal Ed. 93,F153–F161.

Koh, S., Tibayan, F.D., Simpson, J.N., Jensen, F.E., 2004. NBQX ortopiramate treatment after perinatal hypoxia-induced seizuresprevents later increases in seizure-induced neuronal injury.Epilepsia 45, 569–575.

Kohr, G., 2007. NMDA receptor antagonists: tools in neurosciencewith promise for treating CNS pathologies. J. Physiol. 581, 1–2.

Kruger, L., Saporta, S., Swanson, L.W., 1995. Photographic Atlas ofthe Rat Brain. Cambridge University Press, Cambridge, UK.

Lee, S.R., Kim, S.P., Kim, J.E., 2000. Protective effect of topiramateagainst hippocampal neuronal damage after global ischemia inthe gerbils. Neurosci. Lett. 281, 183–186.

Liao, D., Scannevin, R.H., Huganir, R., 2001. Activation of silentsynapses by rapid activity-dependent synaptic recruitment ofAMPA receptors. J. Neurosci. 21, 6008–6017.

Lippert, K., Welsch, M., Krieglstein, J., 1994. Over-additiveprotective effect of dizocilpine and NBQX against neuronaldamage. Eur. J. Pharmacol. 253, 207–213.

Lipton, S.A., 2004. Failures and successes of NMDA receptorantagonists: molecular basis for the use of open-channelblockers like memantine in the treatment of acute and chronicneurologic insults. NeuroRx 1, 101–110.

Lipton, S.A., 2006. Paradigm shift in neuroprotection by NMDAreceptor blockade: memantine and beyond. Nat. Rev. DrugDiscov. 5, 160–170.

Liu, L., Wong, T.P., Pozza, M.F., Lingenhoehl, K., Wang, Y., Sheng,M., Auberson, Y.P., Wang, Y.T., 2004a. Role of NMDA receptorsubtypes in governing the direction of hippocampal synapticplasticity. Science 304, 1021–1024.

Liu, Y., Barks, J.D., Xu, G., Silverstein, F.S., 2004b. Topiramateextends the therapeutic window for hypothermia-mediatedneuroprotection after stroke in neonatal rats. Stroke 35,1460–1465.

Lu, Y.M., Yin, H.Z., Chiang, J., Weiss, J.H., 1996. Ca2+-permeableAMPA/kainate and NMDA channels: high rate of Ca2+ influxunderlies potent induction of injury. J. Neurosci. 16, 5457–5465.

Manning, S.M., Talos, D.M., Zhou, C., Selip, D.B., Park, H.K., Park,C.J., Volpe, J.J., Jensen, F.E., 2008. NMDA receptor blockade withmemantine attenuates white matter injury in a rat model ofperiventricular leukomalacia. J. Neurosci. 28, 6670–6678.

McDonald, J.W., Roeser, N.F., Silverstein, F.S., Johnston, M.V.,1989a. Quantitative assessment of neuroprotection againstNMDA-induced brain injury. Exp. Neurol. 106, 289–296.

McDonald, J.W., Silverstein, F.S., Johnston, M.V., 1989b.Neuroprotective effects of MK-801, TCP, PCP and CPP againstN-methyl-D-aspartate induced neurotoxicity in an in vivoperinatal rat model. Brain Res. 490, 33–40.

Mellon, R.D., Simone, A.F., Rappaport, B.A., 2007. Use of anestheticagents in neonates and young children. Anesth. Analg. 104,509–520.

Micu, I., Jiang, Q., Coderre, E., Ridsdale, A., Zhang, L.,Woulfe, J., Yin,X., Trapp, B.D., McRory, J.E., Rehak, R., Zamponi, G.W., Wang,W., Stys, P.K., 2006. NMDA receptors mediate calciumaccumulation in myelin during chemical ischaemia.Nature 439, 988–992.

Nakajima, W., Ishida, A., Lange, M.S., Gabrielson, K.L., Wilson,M.A., Martin, L.J., Blue, M.E., Johnston, M.V., 2000. Apoptosishas a prolonged role in the neurodegeneration after hypoxicischemia in the newborn rat. J. Neurosci. 20, 7994–8004.

Noh, M.R., Kim, S.K., Sun, W., Park, S.K., Choi, H.C., Lim, J.H., Kim,I.H., Kim, H.J., Kim, H., Eun, B.L., 2006. Neuroprotective effect oftopiramate on hypoxic ischemic brain injury in neonatal rats.Exp. Neurol. 201, 470–478.

Northington, F.J., 2006. Brief update on animal models ofhypoxic–ischemic encephalopathy and neonatal stroke.ILAR J. 47, 32–38.

Northington, F.J., Ferriero, D.M., Graham, E.M., Traystman, R.J.,Martin, L.J., 2001. Early neurodegeneration afterhypoxia–ischemia in neonatal rat is necrosis while delayedneuronal death is apoptosis. Neurobiol. Dis. 8, 207–219.

Palmer, C., Vannucci, R.C., Towfighi, J., 1990. Reduction ofperinatal hypoxic–ischemic brain damage with allopurinol.Pediatr. Res. 27, 332–336.

Pappalardo, A., Liberto, A., Patti, F., Reggio, A., 2004.[Neuroprotective effects of topiramate]. Clin. Ter. 155, 75–78.

Park, H.J., Kim, H.J., Ra, J., Zheng, L.T., Yim, S.V., Chung, J.H., 2008.Protective effect of topiramate on kainic acid-induced celldeath in mice hippocampus. Epilepsia 49, 163–167.

Parsons, C.G., Danysz, W., Quack, G., 1999. Memantine is aclinically well tolerated N-methyl-D-aspartate (NMDA) receptorantagonist—a review of preclinical data. Neuropharmacology38, 735–767.

Poulsen, C.F., Schousboe, I., Sarup, A., White, H.S., Schousboe, A.,2006. Effect of topiramate and dBcAMP on expression of theglutamate transporters GLAST and GLT-1 in astrocytescultured separately, or together with neurons. Neurochem. Int.48, 657–661.

Puka-Sundvall, M., Hallin, U., Zhu, C., Wang, X., Karlsson, J.O.,Blomgren, K., Hagberg, H., 2000. NMDA blockade attenuatescaspase-3 activation and DNA fragmentation after neonatalhypoxia–ischemia. Neuroreport 11, 2833–2836.

Rice III, J.E., Vannucci, R.C., Brierley, J.B., 1981. The influence ofimmaturity on hypoxic–ischemic brain damage in the rat. Ann.Neurol. 9, 131–141.

Rogalewski, A., Schneider, A., Ringelstein, E.B., Schabitz, W.R.,2006. Toward a multimodal neuroprotective treatment ofstroke. Stroke 37, 1129–1136.

Rosi, S., Vazdarjanova, A., Ramirez-Amaya, V., Worley, P.F.,Barnes, C.A., Wenk, G.L., 2006. Memantine protects againstLPS-induced neuroinflammation, restoresbehaviorally-induced gene expression and spatial learning inthe rat. Neuroscience 142, 1303–1315.

182 B R A I N R E S E A R C H 1 2 8 2 ( 2 0 0 9 ) 1 7 3 – 1 8 2

Schubert, S., Brandl, U., Brodhun, M., Ulrich, C., Spaltmann, J.,Fiedler, N., Bauer, R., 2005. Neuroprotective effects oftopiramate after hypoxia–ischemia in newborn piglets.Brain Res. 1058, 129–136.

Sfaello, I., Baud, O., Arzimanoglou, A., Gressens, P., 2005.Topiramate prevents excitotoxic damage in the newbornrodent brain. Neurobiol. Dis. 20, 837–848.

Sinn, D.I., Lee, S.T., Chu, K., Jung, K.H., Song, E.C., Kim, J.M.,Park, D.K., Kim, M., Roh, J.K., 2007. Combined neuroprotectiveeffects of celecoxib and memantine in experimentalintracerebral hemorrhage. Neurosci. Lett. 411,238–242.

Uhm, C.S., Kim, K.B., Lim, J.H., Pee, D.H., Kim, Y.H., Kim, H., Eun,B.L., Tockgo, Y.C., 2003. Effective treatment with fucoidin forperinatal hypoxic–ischemic encephalopathy in rats. Neurosci.Lett. 353, 21–24.

Vannucci, R.C., Vannucci, S.J., 2005. Perinatal hypoxic–ischemicbrain damage: evolution of an animalmodel. Dev. Neurosci. 27,81–86.

Villmann, C., Becker, C.M., 2007. On the hypes and falls inneuroprotection: targeting the NMDA receptor. Neuroscientist13, 594–615.

Volpe, J.J., 2008. Neurology of the Newborn, 5th edn. Elsevier,Philadelphia.

Zhang, G., Dong, Y., Zhang, B., Ichinose, F., Wu, X., Culley, D.J.,Crosby, G., Tanzi, R.E., Xie, Z., 2008. Isoflurane-inducedcaspase-3 activation is dependent on cytosolic calcium andcan be attenuated by memantine. J. Neurosci. 28, 4551–4560.

Zhu, C., Wang, X., Cheng, X., Qiu, L., Xu, F., Simbruner, G.,Blomgren, K., 2004. Post-ischemic hypothermia-induced tissueprotection and diminished apoptosis after neonatal cerebralhypoxia–ischemia. Brain Res. 996, 67–75.

Neuronal glutathione deficiency and age-dependentneurodegeneration in the EAAC1 deficient mouse

Koji Aoyama1,2, Sang Won Suh1,2, Aaron M Hamby1,2, Jialing Liu2,3, Wai Yee Chan1,2, Yongmei Chen1,2 &Raymond A Swanson1,2

Uptake of the neurotransmitter glutamate is effected primarily by transporters expressed on astrocytes, and downregulation of

these transporters leads to seizures and neuronal death. Neurons also express a glutamate transporter, termed excitatory amino

acid carrier–1 (EAAC1), but the physiological function of this transporter remains uncertain. Here we report that genetically

EAAC1-null (Slc1a1–/–) mice have reduced neuronal glutathione levels and, with aging, develop brain atrophy and behavioral

changes. EAAC1 can also rapidly transport cysteine, an obligate precursor for neuronal glutathione synthesis. Neurons in

the hippocampal slices of EAAC1–/– mice were found to have reduced glutathione content, increased oxidant levels and

increased susceptibility to oxidant injury. These changes were reversed by treating the EAAC1–/– mice with N-acetylcysteine,

a membrane-permeable cysteine precursor. These findings suggest that EAAC1 is the primary route for neuronal cysteine

uptake and that EAAC1 deficiency thereby leads to impaired neuronal glutathione metabolism, oxidative stress and

age-dependent neurodegeneration.

Sodium-dependent excitatory amino acid transporters (EAATs) regu-late extracellular glutamate concentrations in the central nervoussystem. Five EAATs have been identified, termed glutamate-aspartatetransporter (GLAST or EAAT1), glutamate transporter 1 (GLT-1 orEAAT2), EAAC1 (EAAT3), EAAT4 and EAAT5 (ref. 1). GLAST andGLT-1 are localized primarily to astrocytes, whereas EAAC1, EAAT4 andEAAT5 are localized primarily to neurons1–4. EAAT4 and EAAT5 arerestricted to cerebellar Purkinje cells and retina, respectively, but EAAC1is widely expressed in neurons throughout the nervous system3,4.

The function of EAAC1 in the brain has not been established. Unlikethe astrocyte glutamate transporters, EAAC1 does not play a major rolein clearing glutamate from the extracellular space5–7. Also unlike theastrocyte glutamate transporters, which are clustered near glutamater-gic synapses, EAAC1 is localized diffusely over cell bodies and pro-cesses2,3,8, suggesting a function other than the re-uptake ofsynaptically released glutamate. An exception to this pattern is pre-synaptic GABAergic terminals, where EAAC1 uptake of glutamatecontributes to re-synthesis of GABA9. An additional distinguishingfeature of EAAC1 is that it can bind and transport cysteine far moreeffectively than the astrocyte glutamate transporters can10,11.

Cysteine is normally the rate-limiting substrate for the synthesis ofglutathione, the principal cellular thiol antioxidant12–14. Most cell typesacquire cysteine in the form of cystine, by hetero-exchange withglutamate (system Xc

–)12, but cell culture studies suggest that neuronslose this capacity during development13,15,16. Consistent with this,studies of the intact mature brain show an absence of Xc

– expression

in neurons17 and extremely low (o 300 nM) cystine concentrations inbrain extracellular fluid18,19. Neuronal glutathione synthesis is sup-ported by astrocytes through an indirect route involving astrocyteglutathione release, its cleavage to cysteinylglycine and the subsequentrelease of free cysteine by an ectopeptidase located on the neuronal cellsurface13,20. The route of cysteine uptake into neurons has not beenascertained, but cell culture studies suggest EAAC1 as a candidateneuronal cysteine transporter because pharmacological inhibitors ofEAAC1 prevent neuronal glutathione synthesis in the presence ofextracellular cysteine14,21,22. Cysteine is transported by EAAC1 at arate comparable to that of glutamate and with an affinity roughlytenfold greater than that of the astrocyte transporters GLAST andGLT-1 (ref. 10).

Glutathione is important for the metabolism of hydrogen peroxide(H2O2), nitric oxide and other reactive oxygen species and for themaintenance of reduced thiol groups on proteins23. Pharmacologicallyinduced glutathione deficiency causes neurodegeneration24, and low-ered glutathione content is found in neurodegenerative disordersassociated with oxidative stress25, suggesting that impairment inneuronal cysteine uptake could lead to neurodegeneration. Here wereport that mice deficient in EAAC1 are deficient in neuronal thiolcontent and develop age-dependent behavioral abnormalities and brainatrophy. Neurons in hippocampal slices from EAAC1–/– mice showedincreased vulnerability to oxidants (but not to glutamate) and reducedcapacity to metabolize reactive oxygen species. Neuronal thiol contentand resistance to oxidant stress were normalized in the EAAC1–/– mice

Received 13 September; accepted 28 October; published online 27 November 2005; doi:10.1038/nn1609

1Department of Neurology, University of California San Francisco, San Francisco, California 94143, USA. 2Veterans Affairs Medical Center, 4150 Clement Street,San Francisco, California 94121, USA. 3Department of Neurosurgery, University of California San Francisco, San Francisco, California 94143, USA. Correspondenceshould be addressed to R.A.S. ([email protected]).

NATURE NEUROSCIENCE VOLUME 9 [ NUMBER 1 [ JANUARY 2006 119

ART ICLES©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

by the administration of N-acetylcysteine (NAC), a membrane-permeable cysteine precursor that does not require active transport.These findings suggest that EAAC1 functions as a neuronal cysteinetransporter and that dysfunction of this system leads to impairedglutathione homeostasis and neurodegeneration.

RESULTS

We determined mouse genotype by polymerase chain reaction (PCR)and confirmed genotype results by western blotting and immunostain-ing for EAAC1 protein expression (Fig. 1). Western blots showedEAAC1 immunoreactivity at the predicted molecular weight(B63 kDa)2,26 in the brains of wild-type mice (Fig. 1b) and noimmunoreactivity in the brains of EAAC1–/– mice. Similarly, immu-nostaining of hippocampal sections showed that EAAC1 was expressedon neuronal cell membranes of wild-type but not EAAC1–/– mice(Fig. 1c). Previous studies have found no change in the expression ofthe major astrocyte glutamate transporters GLT-1 and GLAST inresponse to EAAC1 gene deficiency6 or downregulation7, but expres-sion of these transporters has not been examined in the aged EAAC1–/–

mouse brain. Here, western blots for GLT-1 and GLAST showed nodifference between the wild-type mice and EAAC1–/– mice at either7 weeks (data not shown) or 11 months of age (Fig. 1d), as determinedby densitometry.

Behavioral abnormalities in the EAAC1–/– mice

In maintaining the EAAC1–/– mouse colonies, it became apparent thatthe older mice showed increased aggressiveness and impaired self-grooming compared to age-matched wild-type mice. We studied theirbehavioral changes further with the Morris water maze test27. Theperformance of the wild-type and EAAC1–/– mice was similar at7 weeks of age, showing progressively shortened target latency withrepeated trials on both the visible platform and the hidden platformtasks. By contrast, 11-month-old EAAC1–/– mice did not improve oneither task with repeated trials (Fig. 2a). This impairment was not dueto gross visual disturbances, because the aged EAAC1–/– mice reachednormally for nearby small surfaces when suspended by the tail.Spontaneous locomotor activity was also not significantly altered inthe aged EAAC1–/– mice (data not shown). Spontaneous swim velocitywas slower in the aged EAAC1–/– mice (Fig. 2b), but not slow enough toaccount for the failure to shorten the target latency with repeated trials.Notably, the aged EAAC1–/– mice differed from the other groups in thattheir spontaneous swim speed was well below their maximal swimspeed. Together, these observations suggest cognitive or motivationalimpairment in the aged EAAC1–/– mice.

Brain atrophy and oxidative stress in EAAC1–/– mice

A comparison of coronal sections from wild-type and EAAC1–/– mousebrains showed age-dependent cortical thinning and ventricular enlar-gement in the EAAC1–/– mice. Wild-type mice showed a small increasein ventricular size between the ages of 7 weeks and 11 months (Fig. 3);by contrast, EAAC1–/– mice showed slightly larger ventricle size thanthe wild-type mice at 7 weeks and much larger ventricle size at11 months. Accordingly, measures of the hippocampal CA1 cell layerand the corpus callosum both showed reduced size in the agedEAAC1–/– mice (Fig. 3).

Wild

type

Wild

type

Wild

type

Wild

type

EAAC1–/

–

EAAC1–/

–

EAAC1–/

–

Wild

type

EAAC1–/

–

Wild

type

EAAC1–/

–

EAAC1–/

–

EAAC1EAAC1

β-actin

β-actin

β-actin

(kDa)

170.0

115.5

82.2

64.248.8

Neo

Wild type EAAC1–/–

GLAST

GLT1

a

c

d

b Figure 1 Genotyping and glutamate transporter expression. (a) PCR analysis

of genomic DNA shows loss of the EAAC1 band and presence of the NEO

cassette in the outbred EAAC1–/– mouse. (b) Western blots show the major

EAAC1 band at 63 kDa in the wild-type brain and no immunoreactivity in the

EAAC1–/– mouse brain. (c) Immunostaining of the hippocampal CA1 region

shows EAAC1 expression localized to neuronal cell membranes in the wild-

type brain and no signal from the EAAC1–/– mouse brain. Scale bar, 40 mm.

(d) Western blots for brain GLT-1 and GLAST expression in 11-month-oldwild-type and EAAC1–/– mice.

a

b

Age 7 weeks Age 11 months


60

30

20

10

0

50

40

30

20

10

0

60

50

40

30

20

10

0Day 1

Visible Hidden

Late

ncy

(s)

Sw

im v

eloc

ity (

cm s

–1)

30

20

10

0Sw

im v

eloc

ity (

cm s

–1)

Day 2 Day 3 Day 4 Day 5

Day 1

Visible Hidden


Day 1

Visible Hidden

Late

ncy

(s)


Day 1

Visible Hidden



Wild type EAAC1–/–Figure 2 Performance on the Morris water maze test. (a) At age 7 weeks,

time to reach platform (latency) and rate of latency change was similar in the

wild-type and EAAC1–/– mice during both visible and hidden platform

sessions. For both the visible and hidden platform tasks, the 11-month-old

EAAC1–/– mice showed profound impairment (P o 0.01) as compared to the

7-week- and 11-month-old wild-type mice and the 7-week-old EAAC1–/– mice(n ¼ 10). The dashed and dotted lines indicate the mean rate of change

over the designated testing intervals. (b) Spontaneous swim velocity was

moderately reduced in the 11-month-old EAAC1–/– mice relative to the

7-week-old EAAC1–/– and wild-type mice and the 11-month-old wild-type

mice (P o 0.01, n ¼ 10).

120 VOLUME 9 [ NUMBER 1 [ JANUARY 2006 NATURE NEUROSCIENCE

ART ICLES©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

The age-dependent brain atrophy in the EAAC1–/– mice was accom-panied by markers of oxidative stress. Immunoreactivity for nitro-tyrosine and 4-hydroxy-2-nonenal (HNE), which are formed byoxidant interactions with proteins and lipids, respectively, wasincreased in the aged EAAC1–/– mouse brains. The increase wasprominent in the hippocampal cell fields and the cerebral cortex(Fig. 4). Both nitrotyrosine and HNE localized to neurons in theaged EAAC1–/– mouse brains (Fig. 4). These markers were not detectedin the corpus callosum or other white matter tracts (SupplementaryFig. 1 online).

Increased vulnerability of EAAC1–/– neurons to oxidants

We prepared hippocampal slices from young (6–8 week) wild-type andEAAC1–/– mice to study possible mechanisms by which EAAC1 genedeficiency could lead to the observed age-dependent changes. BecauseEAAC1 can function as both a glutamate and a cysteine transporter10,11,we used hippocampal slices to compare the vulnerability of the wild-type and EAAC1–/– mouse brains to glutamate and oxidant exposures.Neurons in slices from the EAAC1–/– mice did not show increasedvulnerability to glutamate over a range of bath glutamate concentra-tions (Fig. 5). This result is consistent with earlier studies reporting a

ca b

d e f

2 Wild type

Wild type


Wild type

Ven

tric

le a

rea

(mm

2 )

EAAC1–/–

EAAC1–/–EAAC1–/–

EAAC1–/–

EAAC1–/–

EAAC1–/– EAAC1–/– EAAC1–/– EAAC1–/–

EAAC1–/–

Wild type Wild type


Wild type

Wild type

Age 7 weeks

Age 11 monthsWild type

Wild type


Wild type

*

*

**

***

**

1.5

0.5

0

200

150

100

50

0

Str

uctu

re w

idth

(µm

)

Anterior Posterior

7 wee

ks

11 m

onth

s

7 wee

ks

11 m

onth

s

CA1 Corpuscallosum

Corpuscallosum

7 wee

ks

11 m

onth

s

7 wee

ks

11 m

onth

s

1

Figure 3 Brain atrophy in the EAAC1–/– mice. (a,b) Coronal brain sections show cortical thinning and ventricular enlargement in the older EAAC1–/– mice. The

top rows are at the level of the anterior commissure and the bottom rows are 2.7 mm posterior to the anterior commissure. Scale bar, 2 mm in a; 1 mm in b.

(c) Ventricle size is larger in the EAAC1–/– mice at age 7 weeks, and the difference is further increased at age 11 months (n ¼ 6–14). (d) Hematoxylin-eosin

staining of the CA1 hippocampal cell field; scale bar, 40 mm. (e) Fluoro-myelin staining of the corpus callosum (green) in coronal section; scale bar, 100 mm.

Nuclear counterstaining (red) of the pyramidal cell layer caught obliquely on these sections is included for scale and orientation. (f) Quantified measures of the

CA1 cell layer and corpus callosum (n ¼ 3–6). *P o 0.05; **P o 0.01. Error bars denote s.e.m.

a Wild type

Nitrotyrosine

Nitrotyrosine MAP-2 Merge

Nitrotyrosine GFAP Merge

HNE MAP-2 Merge

HNE GFAP Merge

CA

1C

A3

Cor

tex

CA

1C

A3

Cor

tex

EAAC1–/– Wild type

4-hydroxy-2-nonenal

EAAC1–/–b

c

d

Figure 4 Oxidative stress in neurons of EAAC1–/– mouse brain. (a,b) Immunostaining for (a) nitrotyrosine and (b) HNE showed increased immunoreactivity in

neurons of cerebral cortex and hippocampal CA1 and CA3 cell fields in 11-month-old EAAC1–/– mice. Scale bar, 40 mm. (c,d) Immunostaining for nitrotyrosine

and HNE, colocalized with the neuronal marker microtubule-associated protein-2 (MAP-2). Representative of three brains in each group.


ART ICLES©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

negligible role for neuronal glutamate transporters in regulating extra-cellular glutamate concentrations1,9. By contrast, the neurons inEAAC1–/– mice were several times more sensitive to H2O2 and to3-morpholinosydnonimine (SIN-1), which generates superoxide, nitricoxide, peroxynitrite and related oxygen species28.

Cysteine uptake is a rate-limiting step in neuronal synthesis ofglutathione21,23,29, and glutathione has a central role in the metabolismof both peroxide and nitrosyl oxidants23,30,31. SIN-1 greatly increasedneuronal nitrotyrosine immunoreactivity in brain slices fromEAAC1–/– mice under conditions that produced a negligible increasein slices from the wild-type mice (Fig. 6a,b), suggesting an impairedcapacity for the scavenging of nitrosyl radicals in the EAAC1–/– mousebrain31. We performed parallel studies with hippocampal slices fromwild-type mice that had been treated with buthionine sulfoximine(BSO) to reduce brain glutathione content32. Slices from these mice

also showed increased nitrotyrosine formation (Fig. 6a,b), supportingthe possibility that EAAC1 gene deficiency leads to impaired neuronalglutathione synthesis. To further assess this possibility, hippocampalslices from EAAC1–/–, wild-type and BSO-treated wild-type mice wereevaluated with 5-(and 6-)carboxy-2¢,7¢-dichlorodihydrofluoresceindiacetate (DCF), which is oxidized to a fluorescent compound byoxygen species derived from H2O2 or peroxynitrite33. We observed amodest increase in neuronal DCF fluorescence in the wild-type slicesduring incubation with H2O2 or SIN-1 and much larger increases (10-to 40-fold) in the slices from EAAC1–/– mice and BSO-treated wild-typemice (Fig. 6c,d). The effect of BSO pretreatment was slightly less thanthe effect of the EAAC1–/– genotype.

We measured bulk glutathione content in brain homogenates fromEAAC1–/–, wild-type and BSO-treated wild-type mice (Fig. 6e). Thesemeasurements are likely to underestimate the degree of glutathionedeficiency in the EAAC1–/– neurons because a large share of brainglutathione is localized to astrocytes, which do not express EAAC1(ref. 23). Similarly, because the decrease in glutathione in the BSO-treated mice reflects impaired glutathione synthesis in both neuronsand astrocytes, the comparable glutathione reductions observed in theEAAC1–/– mouse brains and BSO-treated mouse brains may indicate amuch greater neuronal glutathione depletion in the EAAC1–/– mice.Glutathione measurements from the livers of wild-type and EAAC1–/–

mice gave comparable values (mean ± s.e.m.)—1.60 ± 0.12 and 1.50 ±0.06 mmol per mg of protein, respectively (n ¼ 4)—suggesting that thereduced glutathione levels in the EAAC1–/– mouse brains results fromlocal rather than systemic effects of EAAC1 gene deficiency.

30

25

20

15

Wild type

Flu

ores

cenc

e in

tens

ity(a

rbitr

ary

units

)

EAAC1–/–

EAAC1–/–

10

5

0

0 h

cont

rol

4 h

cont

rol

4 h

Glu 2.

5 m

M

4 h

Glu 5

mM

4 h

Glu 10

mM

4 h

SIN-1

500

µM

4 h

H 2O2 20

0 µM

Wild type

0 h

cont

rol

4 h

cont

rol

4 h

glut

amat

e 10

mM

4 h

H2O

2 200

µM

4 h

SIN

-1 5

00 µ

M

a b

**

**

Figure 5 Increased vulnerability of neurons in EAAC1–/– mice to oxidative

stress. (a) Neuron death in hippocampal slice preparations was identified by

PI fluorescence. PI fluorescence was evaluated at 0 or 4 h under control

conditions, or 3.5 h after a 30-min incubation with glutamate, hydrogen

peroxide or SIN-1; scale bar, 40 mm. (b) There was a severalfold increased

sensitivity to SIN-1 and H2O2 in the slices from EAAC1–/– mice, but no

increased sensitivity to glutamate. **P o 0.01; n ¼ 3–5. Error bars

represent s.e.m.

Nitrotyrosine

Wild type

**

****

**

**

** **

**50

40

30

20

Arb

itrar

y de

nsity

Arb

itrar

y de

nsity

10

0

60

1.00.80.60.40.2

0

GS

H(µ

mol

per

mg

prot

ein)

40

20

0

DCF fluorescence

0 m

in co

ntro

l

0 m

in co

ntro

l

30 m

in co

ntro

l

30 m

in SIN

-1 5

00 µM

30 m

in H2O

2 20

0 µM

30 m

in co

ntro

l

30 m

in SIN

-1 5

00 µM

Wild type + BSOEAAC1–/–

Wild type

Wild

type

Wild type+ BSO

Wild

type

+ B

SO

EAAC1–/–

EAAC1–/

–

c

30 mincontrol

30 minH2O2 200 µM

30 minSIN-1 500 µM

Wild type Wild type + BSOEAAC1–/–a b

d e

Wild type

0 mincontrol

30 mincontrol

30 minSIN-1 500 µM

Wild type + BSOEAAC1–/–

Figure 6 Reduced scavenging of reactive oxygen species in neurons of EAAC1–/– mice.

(a) Hippocampal slices were prepared from EAAC1–/– mice, wild-type mice or wild-type mice

that had been treated with BSO to reduce brain glutathione content. The slices were evaluated

for nitrotyrosine immunoreactivity after incubation with SIN-1; scale bar, 40 mm. (b) SIN-1

produced a small increase in nitrotyrosine immunoreactivity in the wild-type brain slices

(P o 0.01) and a much larger increase in slices from EAAC1–/– mice and BSO-treated wild-

type mice (**P o 0.01, n ¼ 4). (c) The presence of reactive oxygen species was evaluatedwith DCF after 30-min incubations with H2O2 or SIN-1; bar, 40 mm. (d) Both of the oxidants

produced a small increase in the neuronal DCF signal in the wild-type hippocampal slices

(P o 0.01), and both oxidants produced much larger increases in slices from EAAC1–/– mice

and BSO-treated wild-type mice. (P o 0.01, n ¼ 4). (e) Brain glutathione content was reduced

in the EAAC1–/– mouse brains and in wild-type mice treated with BSO (**P o 0.01, n ¼ 4–8).

Error bars denote s.e.m.


ART ICLES©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

Oxidant effects on neurons are unaffected by bicuculline

The uptake of glutamate by EAAC1 provides a substrate for GABAformation in GABAergic neurons9, raising the possibility that EAAC1deficiency could promote oxidant stress in neurons by reducingGABAergic tone. GABA itself has no significant antioxidant properties,but reduced activation of GABAA receptors could, in principle,indirectly amplify oxidant effects on neurons by increasing neuronaldepolarization, NMDA receptor activation and glutamate release34,35.The comparable neurotoxicity of glutamate in brain slices from wild-type and EAAC1–/– mice (Fig. 5) suggests that this effect, if present,must be small; but to directly test this possibility, we examined theeffect of the GABAA receptor antagonist (+)-bicuculline on theneuronal response to oxidants in the brain slice preparation. Weobserved no effect of 20 mM bicuculline36 on DCF fluorescence inwild-type slices after incubation with SIN-1 or H2O2 (Fig. 7a–d).Bicuculline also had no effect on neuronal survival after incubationwith SIN-1 or H2O2 (Fig. 7e–h).

NAC normalizes neuronal glutathione in EAAC1–/– mice

To determine directly whether EAAC1 is involved in neuronal glu-tathione homeostasis, we used fluorescently tagged C5 maleimide toquantify reactive thiol content (of which glutathione is the principalcomponent) in hippocampal slices from wild-type and EAAC1–/– mice.As expected, the C5 maleimide fluorescence was markedly reduced inneurons from the hippocampal slices of EAAC1–/– mice (Fig. 8a). NACcan passively cross lipid membranes and thereby provide cysteine tocells that lack cysteine transport37,38. The C5 maleimide signal was

normalized in EAAC1–/– mice given NAC 5 h before brain harvest(Fig. 8a). To confirm that this thiol signal was due to glutathione ratherthan to NAC or cysteine, we also prepared slices from EAAC1–/– micethat were given BSO along with NAC to prevent de novo synthesis ofglutathione. The effect of NAC was blocked in these mice (Fig. 8a),confirming that the C5 maleimide signal is primarily attributable toglutathione. Biochemical measures of glutathione further confirmedthat the NAC-induced increase in brain glutathione content wasblocked by BSO and that the BSO treatment did not deplete pre-existing glutathione stores over this time interval (Fig. 8b). Asexpected, neuronal death after oxidant exposure was decreased in slicesfrom NAC-treated EAAC1–/– mice relative to untreated EAAC1–/– mice(Figs. 5 and 8), and this decrease was negated by the coadministrationof BSO (Fig. 8c,d).

DISCUSSION

The original description of the EAAC1–/– mouse reported reducedspontaneous activity but no gross neurodegeneration6. Our studies,using descendants of these mice, similarly showed no gross neuro-degeneration at young ages, but did show brain atrophy and pro-nounced behavioral abnormalities by 11 months of age. These changeswere accompanied by histochemical markers of neuronal oxidativestress, and experiments with acutely prepared hippocampal slicesconfirmed an impaired neuronal resistance to reactive oxygen species.The capacity of EAAC1 to function as a cysteine transporter raised thepossibility that these abnormalities could be caused by impairedneuronal glutathione metabolism, and we confirmed decreased reactive

ca b

e f

14

H2O2

Bic (–)

Bic (+)1210

86420

0 m

in co

ntro

l

30 m

in H 2

O 2 2

00 µM

30 m

in H 2

O 2 1

mM

30 m

in co

ntro

l

Arb

itrar

y de

nsity

d

14

SIN-1

Bic (–)

Bic (+)1210

86420

0 m

in co

ntro

l

30 m

in SIN

-1 5

00 µM

30 m

in SIN

-1 5

mM

30 m

in co

ntro

l

Arb

itrar

y de

nsity

g

12

H2O2

Bic (–)

Bic (+)10

8

6

4

2

0

0 h

cont

rol

0 mincontrol

0 hcontrol

0 hcontrol

4 hcontrol

4 hcontrol

4 hH2O2 200 µM

4 hSIN-1 500 µM

4 hSIN-1 5 mM

4 hH2O2 1 mM

Bic (–) Bic (+)

30 mincontrol

30 minH2O2 200 µM

30 minH2O2 1 mM

0 mincontrol

30 mincontrol

30 minSIN-1 500 µM

30 minSIN-1 5 mM

4 h

cont

rol

4 h

H 2O 2

200

µM

4 h

H 2O 2

1 m

M

Arb

itrar

y de

nsity

h

12

SIN-1

Bic (–)

Bic (+)10

8

6

4

2

0

0 h

cont

rol

4 h

cont

rol

4 h

SIN-1

500

µM

4 h

SIN-1

5 m

M

Arb

itrar

y de

nsity

H2O2

Bic (–) Bic (+)

SIN-1

Bic (–) Bic (+)

H2O2

Bic (–) Bic (+)

SIN-1

Figure 7 Bicuculline does not potentiate the oxidant effects of H2O2 or SIN-1. (a–h) The presence of reactive oxygen species in wild-type hippocampal

cultures was evaluated with DCF after 30-min incubations with H2O2 (a,c) or SIN-1 (b,d) in the presence and absence of 20 mM bicuculline. Neuron death

in wild-type hippocampal slice preparations was assessed by PI fluorescence after incubation with H2O2 (e,g) or SIN-1 (f,h) in the presence and absence of

20 mM bicuculline. Scale bar, 40 mm; n ¼ 4 under each condition. Error bars denote s.e.m.


ART ICLES©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

thiol content in hippocampal neurons of the EAAC1–/– mice. Theadministration of the cell-permeable cysteine precursor NAC correctedboth the neuronal thiol content and neuronal oxidant-scavengingcapacity. The effects of NAC on oxidant-scavenging capacity are dueprimarily to its role as a substrate for glutathione synthesis39,40; weconfirmed this here by showing that the effects of NAC are negated bythe simultaneous administration of BSO.

Immunostaining showed nitrotyrosine and HNE accumulation inneurons in the EAAC1–/– mouse brain, consistent with the localizationof EAAC1 selectively to neurons. The measured loss of neurons in thehippocampal CA1 cell layer was small, however, relative to theincreased ventricular volume. This discrepancy may be explained inpart by associated axonal loss, as demonstrated by the narrowing of thecorpus callosum in the aged EAAC1–/– mice. In addition, there may besecondary loss of glial elements contributing to brain atrophy.

EAAC1 is unique among the cysteine transporters in that itstransport of substrates is coupled to the transport of three Na+ ions,thereby enabling uptake against a steep concentration gradient41. Thepresent findings show that EAAC1 is quantitatively important forneuronal glutathione homeostasis but do not exclude alternative routesof neuronal cysteine uptake that may partially compensate for EAAC1deficiency. In fact, a complete loss of cysteine uptake would not becompatible with prolonged neuronal survival.

EAAC1 is also expressed in several extraneural tissues6. In the kidney,EAAC1 is expressed by renal tubule cells, in which it serves as a majorroute of glutamate and aspartate re-uptake from the urine. EAAC1–/–

mice show dicarboxylic aminoaciduria6, and it is possible that thismetabolic defect could influence neuronal glutathione metabolismindirectly. However, the rapid normalization of neuronal glutathionewith NAC identifies cysteine uptake as the limiting step for glutathionesynthesis in EAAC1–/– mice. Similarly, the normal glutathione contentmeasured in the liver of the EAAC1–/– mouse argues against a systemicdeficiency of sulfur amino acids.

A notable feature of the EAAC1–/– mouse is that the brain atrophyand behavioral abnormalities are markedly age dependent, despite thefact that brain slices obtained from young mice showed large biochem-ical abnormalities when compared to slices from wild-type mice. Thedelayed onset of gross atrophy and behavioral changes may be due to

other interacting factors associated with aging or, alternatively, to theaccumulated effect of a prolonged impairment in neuronal glutathionehomeostasis. In either case, the parallel between the delayed onset ofdisease in these mice and the delayed onset of most human neurode-generative disorders suggests that common mechanisms could beinvolved. There have been case reports of neurological developmentaldelay associated with dicarboxylic aminoaciduria42, but a humangenetic EAAC1 deficiency has not been identified. EAAC1 expressionis, however, regulated by several factors, including protein kinase C43,cholesterol44, presenilin26 and others. It will be of interest to learnwhether the magnitude of these regulatory influences is sufficient toalter neuronal glutathione metabolism.

METHODSStudies were performed in accordance with a protocol approved by the Animal

Use Committee at the San Francisco Veterans Affairs Medical Center. Reagents

were obtained from Sigma-Aldrich except where noted.

EAAC1–/– mice. The EAAC1–/– mice were descendants of the strain established

in a previous study6, in which exon 1 is disrupted by a neomycin resistance

(NEO) cassette. These mice were outbred to wild-type CD-1 mice for six or

more generations before these studies. Littermate and age-matched wild-type

CD-1 mice were used as controls.

Genotyping. Genotypes were confirmed by PCR of tail DNA, using oligonu-

cleotide primers pairs for EAAC1 exon 1 and for the inserted NEO cassette:

5¢-CCGCCACGCAAAACCACCGTGCTCGGTCCC-3¢ (EAAC1.1), 5¢-CTAG

TACCACGGCGGCCACGGTTGAGAGCA-3¢ (EAAC1.2), 5¢-CATTCGACCAC

CAAGCGAAAC-3¢ (NEO.1) and 5¢-CAGCAATGTCACGGGTAGCCAAC-3¢(NEO.2). The amplified products were separated by 1.5% agarose gel electro-

phoresis and visualized by ethidium bromide staining.

Western blotting. Western blotting was performed as described pre-

viously45. Rabbit antibodies to EAAC1, GLAST or GLT-1 (Alpha Diagnostic

International) were used at 0.5 mg ml–1, and mouse monoclonal antibody to

b-actin (Sigma) was used at a 1:1,000 dilution. The antibodies were visualized

by chemiluminescence after incubation with horseradish peroxidase–

conjugated antibody to rabbit IgG. Optical densities of the protein bands

were measured using the NIH ImageJ software program. The protein band

densities were normalized in each case to the density of the b-actin band from

the same sample.

NAC2.0

1.5

1.0

0.5

0

Glu

tath

ione

(µm

ol p

er m

g pr

otei

n)

0

Time (h)

20 ** **

**

EAAC1–/–

+ NAC15

10

5

0

0 h

cont

rol

4 h

cont

rol

4 h

SIN-1

500

µM

4 h

H 2O 2

200

µM

Flu

ores

cenc

e in

tens

ity(a

rbitr

ary

units

)

4 6

BSONAC + BSO

EAAC1–/–

+ NAC + BSO


EAAC1–/– + NACEAAC1–/–

+ NAC + BSO

NAC

EAAC1–/–

NAC + BSO NAC

EAAC1–/–

NAC + BSO

0 hcontrol

4 hcontrol

4 h H2O2200 µM

4 h SIN-1500 µM

a b

d

c

Figure 8 Neuronal glutathione deficiency and vulnerability to oxidant stress in EAAC1–/– mice are both

reversed by NAC. (a) Reactive thiols were evaluated in brain slices using C5-maleimide fluorescence. There

was reduced fluorescence in the neurons in EAAC1–/– slices relative to those in the wild-type slices. This

signal was increased in slices from mice treated with the cell-permeant cysteine precursor NAC, but not in

slices from mice treated with BSO together with NAC to prevent glutathione synthesis. n ¼ 3 in each

group. (b) Biochemical determination of glutathione in EAAC1–/– mouse brains after treatment with NAC or

NAC plus BSO confirmed that the effect of NAC on glutathione levels is attenuated by the concomitant

administration of BSO. *P o 0.05 versus control, n ¼ 3–7. (c,d) PI staining in hippocampal slices

prepared from NAC-treated EAAC1–/– mice showed increased neuronal resistance to H2O2 or SIN-1compared to untreated EAAC1–/– mice (Fig. 5). This effect of NAC was blocked by the coadministration of

BSO. **P o 0.01, n ¼ 3. Error bars denote s.e.m.


ART ICLES©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

Immunostaining. Brain sections were prepared and immunostained as

described45. The primary antibodies were 5 mg ml–1 rabbit antibody to EAAC1

(Alpha Diagnostic International); 1:500 dilution rabbit antibody to nitro-

tyrosine (Chemicon); and 1:500 dilution rabbit antibody to HNE (Alpha

Diagnostic International). After washing, the slices ware incubated with a

1:500 dilution of Alexa Fluor 488–conjugated goat anti-rabbit IgG (Molecular

Probes). Confocal photomicrographs were acquired using 10-nm optical

thickness sections.

Brain morphology measurements. Ventricle area was measured on three

consecutive 30-mm coronal slices taken at the level of the anterior commissure,

and on an additional three slices taken 2.7 mm posterior to the anterior

commissure. Width of the CA1 cell layer in each mouse was averaged from

measurements from three consecutive hematoxylin-eosin–stained slices. Width

of the corpus callosum was measured on three consecutive slices stained with

lipophilic dye fluoromyelin (Molecular Probes) and counterstained with

propidium iodide (PI). Slice images were imported to Photoshop software

for the image analysis, and measurements from the three consecutive slices were

averaged for each data point ‘n’.

Behavioral studies. A total of 40 mice were studied, in two groups of 20. The

first group consisted of 10 wild-type mice and 10 EAAC1–/– mice of age 11–12

months. The second group consisted of 10 wild-type mice and 10 EAAC1–/–

mice of age 6–8 weeks. To minimize the effects of social influences on behavior,

mice were housed individually in the testing room beginning 3 d before testing.

As a first step, spontaneous open field activity was assessed to establish any

locomotor differences. On each of three consecutive days, open field activity

was recorded for 10 min after an initial 1-min adaptation period. Subsequently,

spatial learning and memory were evaluated by the Morris water maze task, as

previously detailed46. The mice were trained first to locate a visible platform

(days 1 and 2), and then to locate a hidden platform (days 3 through 5). The

mice received two training sessions per day, each consisting of three 1-min trials

with a 10-min intertrial interval. Swim speed and time to reach the platform

(latency) was recorded with an EthoVision video tracking system (Noldus

Information Technology). Mice that did not reach the platform within 60 s

were manually placed on the platform and assigned a latency interval of 60 s.

Hippocampal slice preparation. Brains from wild-type and EAAC1–/– mice,

6–8 weeks of age, were vibratome-sectioned into 300-mm coronal slices. The

slices were placed in ice-cold artificial cerebrospinal fluid (aCSF) containing

130 mM NaCl, 3.5 mM KCl, 1.25 mM NaH2PO4, 2 mM MgSO4, 2 mM CaCl2,

20 mM NaHCO3 and 10 mM glucose at pH 7.2 while equilibrated with 95%

oxygen and 5% CO2. The aCSF osmolality was 290 ± 15 mOsm kg–1 as

determined by Wescor vapor pressure osmometer. All studies were done using

brain slices from wild-type and EAAC1–/– mice treated in parallel. Experiments

were initiated by transferring the slices to a circulating bath of aCSF at 30 1C

that was continuously bubbled with 95% oxygen and 5% CO2. Glutamate and

oxidants were added for 30-min incubation intervals. Slices were then trans-

ferred to fresh aCSF and maintained at 22 1C until harvested for cell death or

immunostaining 3.5 h later. Where designated, 20 mM (+)-bicuculline (Tocris)

was incubated with the slices beginning 30 min before the H2O2 or SIN-1

exposures to block GABAA receptor function36,47.

Cell death determinations. Cell death in the brain slices was assessed as

described previously48, with minor changes. Slices were incubated for 3 min

with 5 mM PI, washed, fixed in 4% paraformaldehyde and stored in the dark at

4 1C. Digitized images were prepared with a confocal fluorescent microscope

within 24 h of staining. The PI signal intensity was measured in the CA1 cell

body layer of each slice using a ‘region of interest’ mask of constant size. The

background signal was measured from the stratum radiatum of each slice,

which contained no PI fluorescence, and subtracted from the value obtained in

the CA1 region. Values from three slices per brain were averaged for each ‘n’.

Measurement of reactive oxygen species in hippocampal slices. Hippocampal

slices were pre-incubated in 10 mM 5-(and 6-)carboxy-2¢,7¢-dichlorodihydro-

fluorescein diacetate (DCF; Molecular Probes) for 30 min to allow intracellular

loading. After washing with aCSF, slices were exposed to H2O2 or SIN-1 for

30 min. The slices were immediately photographed to quantify DCF fluores-

cence or fixed in 4% paraformaldehyde for nitrotyrosine immunostaining. DCF

fluorescence and nitrotyrosine immunofluorescence was quantified by the same

method as used for PI fluorescence.

Manipulation and measurement of glutathione level. BSO was injected at a

dose of 660 mg kg–1 i.p. twice a day for 4 d to reduce brain levels of glutathione

in wild-type mice32. NAC was administered as a single 150 mg kg–1 intra-

peritoneal injection 4–6 h before brain harvest to elevate neuronal glutathione

levels in EAAC1–/– mice12,39,49. In some mice, the NAC injection was accom-

panied by BSO (1320 mg kg–1), to prevent de novo glutathione formation.

Biochemical glutathione determinations were performed by the NADPH-

dependent glutathione reductase method, as described previously21. In situ

evaluation of reactive thiols, of which glutathione is the dominant species, was

performed using a fluorescent maleimide derivative50. After incubations under

the designated conditions, hippocampal slices were washed with aCSF, fixed in

4% paraformaldehyde, and incubated overnight at 4 1C with 2.5 mM Alexa

Fluor 488 C5 maleimide (Molecular Probes) in phosphate-buffered saline

containing 2% goat serum, 0.2% Triton X-100 and 0.1% bovine serum

albumin. The slices were photographed with a confocal microscope and the

images quantified as described for PI fluorescence.

Statistics. Data are expressed as means ± s.e.m. The behavioral studies were

assessed using repeated-measures analysis of variance (ANOVA). The brain

morphology and biochemical measurements were assessed by ANOVA and the

Bonferroni test for multiple group comparisons. The fluorescent indicators

used for cell death, immunostaining and thiol determinations were evaluated

with the Kruskal-Wallis test followed by Dunn’s test for multiple comparisons

between the wild-type and EAAC1–/– preparations.

Note: Supplementary information is available on the Nature Neuroscience website.

ACKNOWLEDGMENTSWe thank D. Burns for technical assistance and M. Yenari and S. Massa forcritical reading of the manuscript. This work was supported by grants fromthe US National Institutes of Health and the Department of Veterans Affairs.

COMPETING INTERESTS STATEMENTThe authors declare that they have no competing financial interests.

Published online at http://www.nature.com/natureneuroscience/

Reprints and permissions information is available online at http://npg.nature.com/

reprintsandpermissions/

1. Danbolt, N.C. Glutamate uptake. Prog. Neurobiol. 65, 1–105 (2001).2. Rothstein, J.D. et al. Localization of neuronal and glial glutamate transporters. Neuron

13, 713–725 (1994).3. Shashidharan, P. et al. Immunohistochemical localization of the neuron-specific gluta-

mate transporter EAAC1 (EAAT3) in rat brain and spinal cord revealed by a novelmonoclonal antibody. Brain Res. 773, 139–148 (1997).

4. Arriza, J.L., Eliasof, S., Kavanaugh, M.P. & Amara, S.G. Excitatory amino acid trans-porter 5, a retinal glutamate transporter coupled to a chloride conductance. Proc. Natl.Acad. Sci. USA 94, 4155–4160 (1997).

5. Tanaka, K. et al. Epilepsy and exacerbation of brain injury in mice lacking the glutamatetransporter GLT-1. Science 276, 1699–1702 (1997).

6. Peghini, P., Janzen, J. & Stoffel, W. Glutamate transporter EAAC-1-deficient micedevelop dicarboxylic aminoaciduria and behavioral abnormalities but no neurodegenera-tion. EMBO J. 16, 3822–3832 (1997).

7. Rothstein, J.D. et al. Knockout of glutamate transporters reveals a major role forastroglial transport in excitotoxicity and clearance of glutamate. Neuron 16, 675–686(1996).

8. Coco, S. et al. Non-synaptic localization of the glutamate transporter EAAC1 in culturedhippocampal neurons. Eur. J. Neurosci. 9, 1902–1910 (1997).

9. Sepkuty, J.P. et al. A neuronal glutamate transporter contributes to neurotransmitterGABA synthesis and epilepsy. J. Neurosci. 22, 6372–6379 (2002).

10. Zerangue, N. & Kavanaugh, M.P. Interaction of L-cysteine with a human excitatory aminoacid transporter. J. Physiol. (Lond.) 493, 419–423 (1996).

11. Bendahan, A., Armon, A., Madani, N., Kavanaugh, M.P. & Kanner, B.I. Arginine 447plays a pivotal role in substrate interactions in a neuronal glutamate transporter. J. Biol.Chem. 275, 37436–37442 (2000).

12. Wu, G., Fang, Y.Z., Yang, S., Lupton, J.R. & Turner, N.D. Glutathione metabolism and itsimplications for health. J. Nutr. 134, 489–492 (2004).

13. Dringen, R., Pfeiffer, B. & Hamprecht, B. Synthesis of the antioxidant glutathione inneurons: supply by astrocytes of CysGly as precursor for neuronal glutathione.J. Neurosci. 19, 562–569 (1999).

14. Shanker, G., Allen, J.W., Mutkus, L.A. & Aschner, M. The uptake of cysteine in culturedprimary astrocytes and neurons. Brain Res. 902, 156–163 (2001).


ART ICLES©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

15. Murphy, T.H., Schnaar, R.L. & Coyle, J.T. Immature cortical neurons are uniquelysensitive to glutamate toxicity by inhibition of cystine uptake. FASEB J. 4, 1624–1633(1990).

16. Sagara, J.I., Miura, K. & Bannai, S. Maintenance of neuronal glutathione by glial cells.J. Neurochem. 61, 1672–1676 (1993).

17. Sato, H. et al. Distribution of cystine/glutamate exchange transporter, system x(c)-, inthe mouse brain. J. Neurosci. 22, 8028–8033 (2002).

18. Baker, D.A. et al. Neuroadaptations in cystine-glutamate exchange underlie cocainerelapse. Nat. Neurosci. 6, 743–749 (2003).

19. Wang, X.F. & Cynader, M.S. Astrocytes provide cysteine to neurons by releasingglutathione. J. Neurochem. 74, 1434–1442 (2000).

20. Dringen, R., Gutterer, J.M., Gros, C. & Hirrlinger, J. Aminopeptidase N mediates theutilization of the glutathione precursor CysGly by cultured neurons. J. Neurosci. Res.66,1003–1008 (2001).

21. Chen, Y. & Swanson, R.A. The glutamate transporters EAAT2 and EAAT3 mediatecysteine uptake in cortical neuron cultures. J. Neurochem. 84, 1332–1339 (2003).

22. Himi, T., Ikeda, M., Yasuhara, T., Nishida, M. & Morita, I. Role of neuronal glutamatetransporter in the cysteine uptake and intracellular glutathione levels in cultured corticalneurons. J. Neural Transm. 110, 1337–1348 (2003).

23. Dringen, R. Metabolism and functions of glutathione in brain. Prog. Neurobiol. 62,649–671 (2000).

24. Jain, A., Martensson, J., Stole, E., Auld, P.A. & Meister, A. Glutathione deficiency leadsto mitochondrial damage in brain. Proc. Natl. Acad. Sci. USA 88, 1913–1917 (1991).

25. Schulz, J.B., Lindenau, J., Seyfried, J. & Dichgans, J. Glutathione, oxidative stress andneurodegeneration. Eur. J. Biochem. 267, 4904–4911 (2000).

26. Yang, Y., Kinney, G.A., Spain, W.J., Breitner, J.C. & Cook, D.G. Presenilin-1 andintracellular calcium stores regulate neuronal glutamate uptake. J. Neurochem. 88,1361–1372 (2004).

27. Morris, R. Development of a water-maze procedure for studying spatial learning in therat. J. Neurosci. Methods 11, 47–60 (1984).

28. Hogg, N., Darley-Usmar, V.M., Wilson, M.T. & Moncada, S. Production of hydroxylradicals from the simultaneous generation of superoxide and nitric oxide. Biochem.J. 281, 419–424 (1992).

29. Himi, T., Ikeda, M., Yasuhara, T. & Murota, S.I. Oxidative neuronal death caused byglutamate uptake inhibition in cultured hippocampal neurons. J. Neurosci. Res. 71,679–688 (2003).

30. Hogg, N., Singh, R.J. & Kalyanaraman, B. The role of glutathione in the transport andcatabolism of nitric oxide. FEBS Lett. 382, 223–228 (1996).

31. Beckman, J.S. & Koppenol, W.H. Nitric oxide, superoxide, and peroxynitrite: the good,the bad, and ugly. Am. J. Physiol. 271, C1424–C1437 (1996).

32. Griffith, O.W. & Meister, A. Potent and specific inhibition of glutathione synthesis bybuthionine sulfoximine (S-n-butyl homocysteine sulfoximine). J. Biol. Chem. 254,7558–7560 (1979).

33. Crow, J.P. Dichlorodihydrofluorescein and dihydrorhodamine 123 are sensitive indica-tors of peroxynitrite in vitro: implications for intracellular measurement of reactivenitrogen and oxygen species. Nitric Oxide 1, 145–157 (1997).

34. Avshalumov, M.V. & Rice, M.E. NMDA receptor activation mediates hydrogen peroxide-induced pathophysiology in rat hippocampal slices. J. Neurophysiol. 87, 2896–2903(2002).

35. Nowak, L., Bregestovski, P., Ascher, P., Herbet, A. & Prochiantz, A. Magnesium gatesglutamate-activated channels in mouse central neurones. Nature 307, 462–465(1984).

36. Grover, L.M. & Yan, C. Blockade of GABAA receptors facilitates induction ofNMDA receptor-independent long-term potentiation. J. Neurophysiol. 81, 2814–2822 (1999).

37. Mazor, D. et al. Red blood cell permeability to thiol compounds following oxidativestress. Eur. J. Haematol. 57, 241–246 (1996).

38. Parsons, J.L. & Chipman, J.K. The role of glutathione in DNA damage by potassiumbromate in vitro. Mutagenesis 15, 311–316 (2000).

39. Corcoran, G.B. & Wong, B.K. Role of glutathione in prevention of acetaminophen-induced hepatotoxicity by N-acetyl-L-cysteine in vivo: studies with N-acetyl-D-cysteinein mice. J. Pharmacol. Exp. Ther. 238, 54–61 (1986).

40. Umansky, V. et al. Glutathione is a factor of resistance of Jurkat leukemia cells to nitricoxide-mediated apoptosis. J. Cell. Biochem. 78, 578–587 (2000).

41. Zerangue, N. & Kavanaugh, M.P. Flux coupling in a neuronal glutamate transporter.Nature 383, 634–637 (1996).

42. Smith, C.P. et al. Assignment of the gene coding for the human high-affinity glutamatetransporter EAAC1 to 9p24: potential role in dicarboxylic aminoaciduria and neuro-degenerative disorders. Genomics 20, 335–336 (1994).

43. Gonzalez, M.I., Kazanietz, M.G. & Robinson, M.B. Regulation of the neuronal glutamatetransporter excitatory amino acid carrier-1 (EAAC1) by different protein kinase Csubtypes. Mol. Pharmacol. 62, 901–910 (2002).

44. Canolle, B. et al.Glial soluble factors regulate the activity and expression of the neuronalglutamate transporter EAAC1: implication of cholesterol. J. Neurochem. 88,1521–1532 (2004).

45. Aoyama, K. et al. Acidosis causes endoplasmic reticulum stress and caspase-12-mediated astrocyte death. J. Cereb. Blood Flow Metab. 25, 258–370 (2005).

46. Suh, S.W. et al. Hypoglycemic neuronal death and cognitive impairment are preventedby poly(ADP-ribose) polymerase inhibitors administered after hypoglycemia. J. Neu-rosci. 23, 10681–10690 (2003).

47. Hochman, D.W. & Schwartzkroin, P.A. Chloride-cotransport blockade desynchronizesneuronal discharge in the ‘‘epileptic’’ hippocampal slice. J. Neurophysiol. 83, 406–417(2000).

48. Yin, H.Z., Sensi, S.L., Ogoshi, F. & Weiss, J.H. Blockade of Ca2+-permeableAMPA/kainate channels decreases oxygen-glucose deprivation-induced Zn2+ accumula-tion and neuronal loss in hippocampal pyramidal neurons. J. Neurosci. 22, 1273–1279(2002).

49. Meister, A., Anderson, M.E. & Hwang, O. Intracellular cysteine and glutathione deliverysystems. J. Am. Coll. Nutr. 5, 137–151 (1986).

50. Lantz, R.C., Lemus, R., Lange, R.W. & Karol, M.H. Rapid reduction of intracellularglutathione in human bronchial epithelial cells exposed to occupational levels of toluenediisocyanate. Toxicol. Sci. 60, 348–355 (2001).


ART ICLES©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

Health & Medicine

Education Research: Cognitive performance is preserved in sleep