Effects of Antiepileptic Drugs on Reaction Time, Attention, and Impulsivity in Children

(1)

Effects

of Antiepileptic

Drugs

on

Reaction

Time,

Attention,

and

Impulsivity

in Children

Wendy C. Mitchell, MD*; \‘j Zhou, MS §; John M. Chavez, PhD*; and Bianca L. Guzman*

ABSTRACT. Simple, choice, and complex reaction times,

attention (variability of responses and omission errors), and impulsivity (commission and wrong-hand errors on choice and complex reaction time) were repeatedly mea-sured in 111 epileptic children, aged 5 to 13 years, tested a total of 232 times. Antiepileptic drugs (AEDs) were started, stopped, and adjusted throughout the study

pe-nod, for a variety of clinical indications, and AED serum levels were monitored. The relationship of performance to AED serum level was examined. Overall the

nonspe-cific effect of AEDs was minimal: higher total serum

levels of AEDs correlated with more impulsive errors on complex reaction time testing only. In contrast, in 54 chil-dren receiving carbamazepine monotherapy, we found a dose-related beneficial effect upon reaction time, with higher serum levels associated with faster responses and fewer omission errors, particularly on complex reaction time. Phenobarbital caused minimal dose-related effects: only variability and impulsive errors increased with in-creasing serum levels, and only on one segment of the test (73 subjects). Pediatrics 1993;91:1O1-105; reaction time, anticonvulsants, epilepsy, carbamazepine,

impulsiv-ity,

attention, phenobarbital.

ABBREVIATIONS. AED, antiepileptic drug; PB, phenobarbital; CBZ, carbamazepine; DSCORE, summary drug score.

Antiepileptic drugs (AEDs) are purported to have a variety of adverse effects on behavior and perfor-mance in children.1’2 While general descriptions of AED adverse effects abound in the literature, docu-mentation of specific dose-dependent change in cog-nitive function is often lacking, particularly in chil-dren. It is unclear whether most cognitive effects of AEDs are dose dependent or idiosyncratic. Clinical descriptions of children taking AEDs as “hyperac-tive,” “less attentive,” and “dull” have rarely been

supported by objective comparisons with

age-matched children taking other medications.

Phenobarbital (PB) has been reported to have sub-tle adverse effects on cognitive function, at least in infants and toddlers. Farwell et a13 found a signifi-cant decrease in IQ in toddlers treated for 2 years with PB for febrile seizures. Camfield et al4 found detrimental effects of PB on memory in a similar group of children with febrile seizures. However,

From the *Neurology Department, Childrens Hospital of Los Angeles; and

Departments of jNeurology and §Preventive Medicine, Biometry Program.

University of Southern California School of Medicine, Los Angeles. Received for publication Dec 10, 1991; accepted Jun 19, 1992. Presented in part to the American Epilepsy Society. 1988.

Reprint requests to (W,G.M.) Neurology. Mail Stop 82, Childrens Hospital

Los Angeles, P0 Box 54700, Los Angeles. CA 90054.

Acad-emy of Pediatrics.

Hara5 compared a group of unmedicated or

PB-treated children with febrile or afebrile seizures with a group of normal children by using a computerized test of attention. He found that children with seizures differed from normal subjects, regardless of AED treatment.

Carbamazepine (CBZ) is often believed to have fewer adverse effects than sedative AEDs such as PB

and primidone, and some authors report that CBZ

improves specific cognitive functions, in a dose-re-lated manner. Schain et al6 reported an improvement in cognitive function in children who took CBZ in place of one or more AEDs in an open-label study. Carbamazepine had a positive effect on tasks requir-ing a reflective, analytic cognitive style, but no change in overall performance on intelligence tests. Their results, however, are biased by the selection of subjects doing poorly on other AEDs for crossover to

CBZ, and by lack of blinding. In a small, randomized,

double-blind study comparing PB and CBZ in chil-dren with new-onset partial seizures, we found no differences in overall cognitive function or parental reports of behavioral changes.7

Specific concentration-dependent cognitive effects of CBZ have been studied by several investigators. Aman et a!8 studied the relationship of CBZ dose to psychomotor performance by testing children either after a scheduled dose of CBZ or delaying the dose until after the test, to compare performance at the

time of peak vs trough CBZ concentration. They

found some positive concentration-dependent effects of CBZ: higher CBZ level was associated with less movement in the seat (fidgeting), fewer commission errors on a continuous performance test, but incon-sistent changes in response times. In general, delay-ing the medication dose (testing at a lower serum concentration) led to poorer performance. O’Dough-erty et al,9 using several neuropsychological mea-sures including a computerized continuous perfor-mance task, studied 11 children treated with CBZ. Subjects were studied before starting CBZ and at low or moderate serum levels at two subsequent sessions. Serum levels were varied by changing the time be-tween the CBZ dose and the test session, rather than by changing the total dose. Higher CBZ level was

associated with improved manual dexterity and

speeded eye-hand coordination, but increased errors on paired-associates learning.

We previously reported that intellectually normal children with epilepsy were significantly slower than normal children on simple, choice, and complex re-action time. They were more variable in their re-sponses but did not make more errors of omission

at Viet Nam:AAP Sponsored on September 1, 2020

www.aappublications.org/news

(2)

8.3 4.5-13.5 49 25 21 3 2 60 38 2 62 32 15 46 27 8 19

* Total greater than 100% due to 9% with mixed seizures. 1 Includes 7% with cysticercosis.

(measures of attention), and they made more impul-sive errors only on the most complex task.1#{176}Severity and duration of seizure history did not affect these functions. Total medication exposure (calculated to account for number of anticonvulsants used and total time of use) was related to complex reaction time, variability, and omission errors on complex reaction time test and to omission errors on choice reaction time. We found small but statistically significant dif-ferences between subjects with less than I month of total medication exposure and those receiving AEDs for more than a year. We continued to study the same group of subjects, as AEDs were begun, discontin-ued, or changed and as doses were adjusted. We now report the results of these longitudinal investigations. We asked whether there is a dose-dependent effect of AEDs in general or a specific effect of either CBZ or PB on reaction time, impulsivity, or attention.

METHODS

We developed a test formatted as a simple video-game. The test consists of three segments, described to the child as games, pro-gressing in difficulty: simple reaction time for each hand, choice reaction time, and complex reaction time (choice with distracting stimulus). Development and standardization of the test has been described in detail elsewhere.10-11 The scores included three motor speed measures: simple, choice, and complex reaction time, each recorded for dominant and nondominant hand. Measures of at-tention include variability of responses on each segment and er-rors of omission; measures of impulsivity include wrong-hand and commission errors on choice and complex reaction time tests. Sub-jects were children with seizures enrolled in a longitudinal study of cognitive, behavioral, and psychosocial aspects of childhood epilepsy. Subjects were of diverse ethnic backgrounds and socio-economic levels, English or Spanish speaking, and without motor or sensory deficits that could interfere with testing. Demographic and diagnostic information is described in Table I. Moderately and severely retarded children were not enrolled. Testing was usually done on the day of a regular clinic visit, occasionally on a visit specifically for psychometric testing. Testing was omitted if the family reported convulsions in the previous 24 hours.

Antiepileptic drug treatment was prescribed and adjusted by the child’s physician. Subjects were given a variety of medications, either monotherapy and polytherapy. Medications and doses were changed for clinical indications, including lack of seizure control,

TABLE 1. Demographic and Diagnostic Characteristics of

Sub-jects Age, y Median Range Ethnicity, % Hispanic African-American Non-Hispanic White Asian

Other or mixed Primary language, %

English Spanish Other

Seizure types,* % Convulsions

Partial (simple, complex) Absence

Etiology, % Unknown

Age-related epilepsies of childhood Primary generalized

Benign focal epilepsy Remote symptomatict

adverse effects, or discontinuation of treatment after a period of full seizure control. No attempt was made to standardize drug therapy as part of the study. Serum AED levels were requested as part of routine clinical care and were measured by the clinical laboratories using fluorescence polarization immunoassay on a Tdx System (Abbott Laboratories, North Chicago, IL). Blood was generally drawn either just prior to or just after performance of the video-game test, always within an hour. Visits on which no sera were drawn were excluded, unless no AED was prescribed. Chil-dren treated with medications not generally monitored with se-rum levels (primarily benzodiazepines) were excluded from the study.

Summary Drug Score

A summary drug score was used to look for effects of all med-ications combined. Summary drug score (DSCORE) was calcu-lated as shown in Table 2, comparing AEDs using the generally accepted therapeutic range of serum concentrations ie, I = below the usual therapeutic concentration, 2 = lower third of the thera-peutic range, 3 = middle third, 4 = upper third, and 5 = serum levels above the usual therapeutic range). For patients receiving polytherapy, DSCORE was the sum of the individual drug scores. For analyses using DSCORE, all test sessions of each subject were used.

Psychometric Testing

Intelligence was tested using either the Full-Scale IQ from the Wechsler Intelligence Scale for Children-Revised, the General Cog-nitive Index from the McCarthy Scale of Children’s Abilities, or the Full-Scale IQ from the Escala Inteligencia por Ninos Wechsler-Revisada for Spanish-speaking subjects. The Full-Scale IQ and the General Cognitive Index both have a mean of 100 but have slightly different standard deviations. They are all well-standardized tests, validated in appropriate populations, and comparable for overall assessment of cognitive function.52-53 For purposes of this study, they were considered equivalent.

Data Analysis

Individual game scores were transferred to an IBM 3081 corn-puter. Statistical analysis was performed using SAS (SAS Institute, Cary, NC). A median reaction time (MRT) for each subject for each test was calculated from the raw reaction times. The mean devia-tion of the response times (MDRT) was used as a measure of variability of responses. It was calculated using the following for-rnula:

sum ) RT-MRT(

MDRT= N

Reaction times and variability scores were transformed to natural logs and entered as the dependent variables in multiple regres-sions, with either serum AED level or DSCORE entered as inde-pendent variables. IQ and age were entered as independent van-ables in analyses of reaction time and variability, but not in analyses of error scores, which were previously found to be un-related to both age and IQ. SAS General Linear Models procedures were used initially entering age, IQ, and either AED level or DSCORE, and interaction of age and AED level or DSCORE. In-teraction term was never significant and was dropped from later analyses.

Pairs of test sessions were examined for concentration-depen-dent intraindividual differences. To look at total drug effect, pairs of sessions were categorized as increasing AED (total DSCORE increase of 2 or more), decreasing AED (total DSCORE decrease of

2 or more), and no change (DSCORE within I point). Analysis of variance was used. Pains of sessions with no change in DSCORE were used to examine overall practice effect. Age was not entered in the analysis of change in DSCORE, as pairs of tests were never more than I year apart, an interval over which age has little effect. Similarly, IQ was measured only at entry and was assumed to be stable between testing sessions.

To examine effect of change in serum level of each individual AED, serum level from the first session was subtracted from that at the second session of the pain. Change in reaction times,

van-ability, and errors between sessions was calculated by subtracting

score from the first session from the score at the second. Change in AED level was correlated with change in reaction times, vaniabil-ity, and errors, treating both change in drug level and change in test score as continuous vaniables.

(3)

TABLE 2. Summary D rug Sconesv

Scone

‘‘‘L

CBZ Phenytoin Ethosuximide Valpnoic Acid

01- .. . .-. ... ... --

-1 <15 <4 <10 <40 <40

2 15-23 4-6 10-13 40-60 40-60

3 24-32 7-9 14-17 61-80 61-80

4 33-40 10-12 18-20 81-100 81-100

5 >40 >12 >20 >100 >100

* All serum levels in micrograms pen milliliter, rounded to whole digit. Patients neceiving pnmidone monotherapy scored by phenobanbital (PB) level. CBZ, canbamazepine.

t No drug detected; none prescribed.

All results are reported with uncorrected significance values (P

value not corrected for multiple comparisons).

RESULTS

A total of 232 tests of 111 individuals were avail-able to assess total drug effect, representing all

mdi-viduals on AEDs with available measurements of

serum levels at the time of the test, as well as subjects tested while taking no AEDs. Individual subjects were tested one to five times.

IQ scores ranged from 69 to 117, with a median of

91 . Age and IQ were strongly negatively correlated

with each reaction time and variability score but were not significantly related to error scores.

Effects of All AEDs Combined (DSCORE)

Multivariate Analyses (All Test Scores, All Individuals, All AEDs). After controlling for IQ and age, total drug score (DSCORE) was not significantly related to any of the reaction time or variability scores. Only commission and hand errors on complex reaction time testing were positively correlated with total drug score, at statistically significant levels (see Table

3).

Paired Examinations (Intraindividual Pairs, All AEDs). Pairs of examinations were used to look at

the relationship between change in DSCORE and

change in test scores. One hundred eighteen paired scores were available, including subjects in whom one test was performed with no AED, either prior to starting treatment or after tapering medications.

Practice effect between sessions was minimal, never reaching statistical significance. Total DSCORE in-creased (+2 or more) in 35, decreased (-2 or less) in

38, and stayed the same (-1, 0, or +1) in 45 paired test sessions. Change in DSCORE correlated with change in variability on simple reaction time (in-creasing variability with increasing DSCORE), and with commission and hand errors on complex reac-tion time (more errors with increasing DSCORE). All other correlations were statistically insignificant (see Table 4).

Effect of CBZ

Multivariate Analyses (All Tests for Individuals Re-ceiving CBZ Monotherapy). Seventy-nine tests were available for 54 individuals (32 female, 22 male) re-ceiving CBZ monotherapy (Table 3). Carbamazepine levels ranged from subtherapeutic (2 pg/mL) to

above the usual therapeutic range (13 pglmL). As

expected, age was a significant predictor of all reac-tion time and variability scores. IQ was negatively associated with simple but not complex reaction time scores, and with variability scores. After controlling for IQ and age, CBZ level was negatively correlated with the following variables at statistically signifi-cant levels (P < .05): simple reaction time, nondom-inant hand only; complex reaction time for each hand; all variability scores (on simple, choice, and complex reaction time). Omission errors were nega-tively correlated with CBZ level for choice and com-plex reaction time. Commission error scores were

TABLE 3. Significant Relationships Between Antiepileptic Drug L evel and Test Score*

Scone Other Variables

Entered

Direction of Effect

P Value

For DSCORE (n 111 individuals, 232 tests)

Hand enrons, complex RT Age, IQ Positive <.03

Commission errors, complex RT Age Positive <.005

For CBZ (n = 54 individuals, 79 tests)

Simple RT, nondominant Age, IQ Negative <.04

Complex RT, dominant Age, IQ Negative <.008

Complex RT, nondominant Age, IQ Negative <.05

Variability, simple RT Age, IQ Negative <.01

Variability, choice RT Age, IQ Negative <.02

Variability, complex RT Age, IQ Negative <.02

Omission errors, choice RT Age, IQ Negative <.01

Omission errors, complex RT Age, IQ Negative <.03

For PB (n = 73 individuals, 103 tests)

Variability, choice RT Age, IQ Positive <.03

Omission errors, complex RT Age Positive <.04

Choice RT, dominant Age, IQ Positive NS, P = .053

Choice RT, nondominant Age, IQ Positive NS, P = .06

* IQ either Wechslen Intelligence Scale for Children-Revised Full-Scale IQ or McCanthy General Cognitive Index, normalized; RT, reaction time; DSCORE, summary drug scone; CBZ, carbamazepine; PB, phenobarbital; NS, not significant.

(4)

TABLE 4. Paired Test s: Speanman Correlation s of Change in D rug Level to Change in Score*

Score

CBZ (n - 38 Pairs)

r P

DSCORE (n = 118 Pains) PB (n = 43 Pairs)

r P

Simple reaction time

Dominant -.11 NS +004 NS +08 NS

Nondominant -.04 NS +09 NS +11 NS

Variability +21 NS +24 .008 +17 NS

Omission errors -.20 NS +03 NS +02 NS

Choice reaction time

Dominant +02 NS +003 NS +07 NS

Nondominant -.07 NS +05 NS +08 NS

Variability -.23 NS -.04 NS -.01 NS

Omission errors -.49 .002 -.13 NS -.06 NS

Hand errors -.12 NS -.07 NS -.06 NS

Complex reaction time

Dominant -.37 .05 +09 NS +11 NS

Nondominant -.41 .03 +08 NS +15 NS

Variability -.40 .04 +04 NS -.04 NS

Omission errors +23 NS +19 NS +22 NS

Hand errors +32 .09 +23 .02 +21 NS

Commission errors +35 .06 +22 .03 +26 NS

*CBZ, canbamazepine; DSCORE, summary drug scone; PB, phenobanbital; NS, not significant.

positively correlated with CBZ level for the most complex tasks but did not reach statistical signifi-cance.

Paired Tests (Individuals Receiving CBZ Monotherapy on Two or More Occasions, or Starting or Stopping CBZ, With One of Pair Without AED Treatment). Paired tests were available for 38 subjects receiving CBZ mono-therapy. Pairs included subjects in whom CBZ was started or stopped, with one of the sessions done in an untreated state. Drug level was unchanged in 17 (level within ±2 pg/mL between the two sessions), decreased in 9, and increased in 12. Table 3 details the correlations between change in CBZ level and change in test score. The following variables were

signifi-cantly negatively correlated with change in CBZ

level: change in complex reaction time for each hand, variability of complex reaction time, and omission errors on choice reaction time. Negative correlation between change in score and change in serum level implies that when CBZ concentration increased, re-action time or error rate decreased for that individ-ual. There was a trend toward increased impulsive errors on complex reaction time testing as level went up, but this did not reach statistical significance.

Effect of PB

Multivariate Analyses (All Subjects Receiving PB Monotherapy). One hundred three tests were

avail-able for 73 subjects receiving PB monotherapy (Table 3). Phenobarbital serum levels ranged from subther-apeutic (3 pg/mL) to the upper-third of the usual therapeutic range (34 pg/mL). None were above the usual therapeutic range or clinically toxic. Phenobar-bital serum level was significantly positively corre-lated with variability on choice reaction time, after controlling for age and IQ, and omission errors on complex reaction time. Although reaction time mea-sures all were positively correlated with PB level, none of these associations reached statistical signifi-cance.

Paired Tests (Individuals Receiving PB Monotherapy on Two or More Occasions, or Starting or Stopping PB,

With One of Pair Without AED Treatment). Forty-three pairs were available: 11 increased, 13 decreased, and

19 unchanged. Pairs included subjects in whom PB

was started or stopped, with one of the tests per-formed in an untreated state. No significant correla-tions of change in score with change in drug level were found (Table 4).

Although subjects taking a variety of other AEDs were included in the analysis of DSCORE, no other

group receiving monotherapy was large enough to

analyze the individual effects of AEDs other than

phenobarbital and carbamazepine. Similarly, not

enough subjects were available to analyze specific interactions between seizure types, drug, or drug level.

DISCUSSION

Antiepileptic drugs are generally thought to have only adverse effects on behavior and attention. In this study, CBZ consistently improves performance on reaction time and variability measures, in a serum level-related manner. Although the effect was not statistically significant for all measures, there are no reaction time or variability measures worsened by

CBZ, as might be expected if the drug had no effect,

or the results represent chance variation. There is a slight but significant CBZ serum concentration-re-lated increase in impulsive errors on the most com-plex test, but not on simple or choice reaction time. When we combine data for all AEDs as both mono-therapy and polytherapy, we find a striking lack of consistent serum concentration-related effect on re-action time, attention, or impulsivity. Only one mea-sure of impulsivity (hand and commission errors on the most complex task) has a significant concentra-tion-related relationship with total AEDs. Among children receiving PB monotherapy, there is a trend toward slower reaction times and increased omission errors as PB levels increase, but none of the relation-ships reach statistical significance, and the changes are very small. However, there are relatively few sub-jects with PB levels in the upper third of the usual therapeutic range. The apparently opposite effects of

CBZ and PB on reaction time and variability may

explain the lack of overall effect of total drug expo-sure.

(5)

There are several limitations that must be kept in mind in interpreting this study: patients were not randomized either to specific AEDs or to changes in AED dosages. Antiepileptic drugs and dosages were determined clinically. Thus, we cannot compare all subjects receiving PB to all subjects receiving CBZ,

since we cannot rule out the possibility that the choice of AED was made on the basis of behavioral differences or prior history of adverse response to AEDs. However, we can make intrasubject compari-sons as AED levels changed, regardless of reason for change in level. We believe that this approach is valid, despite the many possible reasons for change in AED level: increasing dosages to improve seizure control, decreasing level due to noncompliance, de-creasing dosage due to prolonged period free of sei-zures, or alteration in dosages due to adverse effects of AED, for example. Although seizure control may have been different on two successive test sessions, we found that neither seizure frequency nor severity influenced test scores.1#{176}Clinical judgment may have moderated the effects we have seen and may be re-sponsible for the lack of substantial adverse effects of PB or AEDs examined together. If a child taking an

AED was perceived to have adverse behavioral

ef-fects, the dosage may have been decreased, or further increases avoided. However, we think it is quite un-likely that these clinical assumptions account for the apparent beneficial effects of CBZ on attention and reaction time, which were present in both the CBZ group as a whole and in intrasubject comparisons for those tested two or more times while taking CBZ.

Carbamazepine is structurally related to the tricy-clic antidepressants and has been reported to have antidepressant properties. Improved behavior and mood were reported in several early studies of CBZ as an anticonvulsant.6’14’15 However, mood and be-havioral improvements in early studies generally re-flected the substitution of moderate doses of CBZ for large doses of other AEDs and may have been due to selection bias, as patients doing well while receiving other AEDs may not have been entered in the study.

Stimulant drugs improve performance on tests of reaction time and attention. Tricyclic antidepressants, imipramine in particular, may also improve cogni-tive performance in children. Werry et al16 compared imipramine and methylphenidate in hyperactive children, using a battery of tests including a

contin-uous performance test, and found improvement in

response times and errors with both drugs, although the effect was generally larger with methylphenidate. Carbamazepine is pharmacologically similar to imi-pramine and may share some of its beneficial prop-erties with respect to attention and motor speed.

We conclude that there is no general effect of AEDs on reaction time, attention, and impulsivity. We find surprisingly little serum concentration-dependent ef-fect of PB to the middle of the therapeutic range but recognize that the clinical tendency to use only low

to moderate dosages of PB may have limited the

observed effects. In contrast, CBZ appears to have a beneficial effect on reaction time and attention, with minimal cost in terms of increased impulsivity. Our results generally agree with those of Aman et al,8 who used substantially different tests and a different

study design. The minimal effects found combining all AEDs may reflect the fact that while some AEDs have a slightly detrimental effect, CBZ, given to ap-proximately one third of our subjects, appeared to have a mildly beneficial effect.

Despite the apparent serum level-related beneficial effect of CBZ on reaction time, we do not recommend that CBZ dosages be increased to maximize serum

level in a child whose seizures are well controlled

with low serum levels. The reported tests evaluate only limited aspects of cognitive and behavioral functioning. We do not currently recommend the use of preferential use of CBZ in childhood epilepsy if other AEDs have successfully controlled seizures with minimal perceived adverse effects. Any benefi-cia! effects of CBZ on other conditions such as atten-tion deficit are purely speculative at this point and should be the subject of future research.

ACKNOWLEDGMENTS

This study was funded by a grant from The Robert Wood Johnson Foundation.

We thank Dr Stanley Azen, University of Southern California School of Medicine, Department of Preventive Medicine, for his

supervision of data analysis and biostatistics. We also thank Dr

Masato Takihashi, Children’s Hospital, for his translation of Ha-na’s article from Japanese. Computer programming and assistance in development of the computerized tests was provided by Dr Stephen Citron, Professor of Electrical Engineering, Purdue Uni-versity.

REFERENCES

1. Vining EPG, Mellits ED, Dorsen M, et al. Psychologic and behavioral effects of anticonvulsant drugs in children: a double blind comparison of phenobarbital and valproate. Pediatrics. 1987;80:165-174

2. Trimble MR. Antiepileptic drugs. cognitive function and behavior in

children: evidence from recent studies. Epilepsia. 1990;31(suppl):S30-534 3. Farwell JR, Lee YJ, Hirtz DC, Sulzbacher SJ, Ellenberg JH, Nelson KB.

Phenobarbital for febrile seizures: effects on intelligence and on seizure

recurrence. N Engi IMed. 1990;332:364-369

4. Camfield CS, Chaplin S. Doyle AV, Shapiro SH. Cummings C. Camfield PR. Side effects of phenobarbital in toddlers: behavior and cognitive aspects. IPediatr. 197995:361-365

5. Hara H. Sustained attention and the influence of anticonvulsants on it in children with epilepsy or febrile convulsions. No To Hattatsu. 1986;18:387-398

6. Schain RJ, Ward JW, Guthrie D. Carbamazepine as an anticonvulsant in

children. Neurology. 1977;27:476-480

7. Mitchell WG, Chavez JM. Carbamazepine versus phenobarbital for par-tial onset seizures in children. Epilepsia. 1987;28:5&-60

8. Aman MF, Werry JS, Paxton JW, Turbott SH, Stewart AW. Effects of carbamazepine on psychomotor performance in children as a function of drug concentration, seizure type and time of medication. Epilepsia.

1990;31:51-60

9. O’Dougherty M, Wright FS. Cox S. Walson P. Carbamazepine plasma concentration: relationship to cognitive impairment. Arch Neurol.

1987;44:863-867

10. Mitchell WG, Chavez JM, Zhou Y, Guzman B. Reaction time, attention

and impulsivity in children with epilepsy. Pediatr Neuro!. 1992;8:19-24

11. Mitchell WG, Chavez JM, Baker SA, Guzman BL, Azen SP. Reaction time, attention and impulsivity in normal and hyperactive children.

JChild Neurol. 1990;5:195-204

12. Arinoldi CG. Concurrent validity of McCarthy’s Scales. Percept Mot Skills. 1982;54:1343-1346

13. Rodriguez VJ, Prewitt Diaz JO. Correlations among GPA and scores on the Spanish version of the WISC-R and the Woodcock-Johnson Achieve-ment subtests for 10- to 12-year-old Puerto Rican children. Psychol Rep.

1990;66:563-566

14. Thompson PJ. Trimble MR. Anticonvulsant drugs and cognitive func-tions. Epilepsia. 1982;23:531-544

15. Dodrill CC, Troupin AS. Psychotropic effects of carbamazepine in

epi-lepsy: a double-blind comparison with phenytoin. Neurology. 1977;27:1023-1028

16. Werry JS. Aman M, Diamond E. Imipramine and methylphenidate in hyperactive children. IClxild Psychol Psycixiatry. 1980;21:27-35

(6)

1993;91;101

Pediatrics

Wendy G. Mitchell, Yi Zhou, John M. Chavez and Bianca L. Guzman

Children