The Neuropsychological effects of valproate with and without carnitine supplement in a pediatric migraine sample


Material Information

The Neuropsychological effects of valproate with and without carnitine supplement in a pediatric migraine sample
Physical Description:
vi, 109 leaves : ; 29 cm.
Fiano, Kristin M., 1965-
Publication Date:


Subjects / Keywords:
Migraine -- Child   ( mesh )
Migraine -- Infant   ( mesh )
Migraine -- drug therapy   ( mesh )
Valproic Acid -- therapeutic use   ( mesh )
Valproic Acid -- pharmacology   ( mesh )
Valproic Acid -- adverse effects   ( mesh )
Carnitine -- therapeutic use   ( mesh )
Carnitine -- pharmacology   ( mesh )
Carnitine -- adverse effects   ( mesh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1994.
Includes bibliographical references (leaves 100-108).
Statement of Responsibility:
by Kristin M. Fiano.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 50408786
System ID:

Full Text








The amount of time and effort required to conduct a study

and compile a document of this magnatude is enormous and could

not have been done without the exhaustive efforts of many

people. The author would like to extend her sincerest

gratitude to the members of her committee who ensured that

this effort was not wasted. They have provided candid,

constructive feedback at every point and are congratulated on

their superb work as committee members. Special thanks go to

a non-committee member, Paul Kubilis of Biostatistics, for his


The author would also like to make a special

acknowledgement to her committee chairperson, Eileen Fennell,

who has filled many roles throughout the past five years. She

has served as advisor, mentor, and friend, and has left a

lasting impression on the author and her future work. Her

care and expertise are deeply appreciated.

Finally, personal appreciation is extended to family

members, particularly the author's husband Ric Fiano, who

prevented the world from falling apart on several occasions,

and to Edward and Patricia Galloway, the author's parents,

whose love and pride finally paid off.



ACKNOWLEDGEMENTS ...................

ABSTRACT.......... .................


1 INTRODUCTION ..............

2 METHODS ....................

Materials and Apparatus....
Hypotheses .................

3 RESULTS ....................

4 DISCUSSION .................

Interpretation of Findings.
Adolescent Noncompliance...
Limitations of Study.......






: : : : : : :

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy




August 1994

Chair: Eileen B. Fennell
Major Department: Clinical and Health Psychology

Adult studies of patients with migraine suggest that some

patients have difficulties in verbal memory and attention.

Little valid research has been conducted in this area on a

pediatric sample. This study assessed the presence of

neuropsychological differences in children with migraine

compared to normals, and the effect of a medication used in

the treatment of migraine.

Whether an antiepileptic drug used for the prophylaxis of

migraine has a detrimental effect on neuropsychological

processes is also an area of controversy. Valproate, used in

this study, has historically been purported to exert fewer

negative side-effects on cognition than other antiepileptic

medications, (e.g., phenobarbital). However, such research

has primarily used epilepsy patients, thus confounding the

effect of diagnosis with drug effect. The purpose of this

study was to assess the cognitive effects of valproate in

children without a diagnosis of epilepsy.

Fifteen patients recruited from the pediatric clinic at

Shands Hospital were enrolled in the study. Twenty control

subjects, similar in age and gender distribution, were

recruited from a local school associated with the University

of Florida and from a local church group. All subjects were

administered a neuropsychological battery at three testing

points. At baseline, intellectual screening was used to

assure a normal range of IQ. Measures included verbal memory,

nonverbal memory, memory span, attention, motor tapping, and

parent reports of problem behavior. Migraine subjects

received valproate monotherapy for one month after baseline

testing before their second testing. Subjects were then

divided into valproate plus carnitine supplement, valproate

plus placebo, and placebo plus placebo groups, for an

additional eight weeks. A third testing was administered at

that time. Control subjects were administered the same tests

in the same manner, but without treatment.

Children and adolescents with migraine were not found to

differ from normals at baseline. Once migraine therapy was

initiated, modest declines in attentional functioning were

observed but problems with attention did not impact memory.

Another finding was the significantly higher rate of

internalizing problem behaviors among the children with

migraine, which is consistent with some findings in the adult

migraine literature. Limitations and further interpretations

are discussed.



Headaches in the general population are quite common

and, due to an estimated 150 million lost workdays, can pose

problems for the work force in this country. Approximately

90 percent of the United States population are affected by

headache, with as many as 45 million forced to contend with

headache pain (Silberstein, 1991). While headache pain for

many people is often tension related, others experience

headaches associated with the phenomenon of migraine. This

subset of headache may also have stress as a trigger, but

the mechanism is thought to be different than in the common

tension headache, and has an added hereditary component.

Historically, adults are the first to be eligible for

newly emerging medical techniques or to be the focus of

research. Gradually, however, more research is being

dedicated to these problems in children. Although children

do not readily come to mind when discussing tension

headaches or migraine, they do in fact represent a

substantial percentage of the general population affected by

such a condition (Fenichel, 1985). As a result, migraine

headaches in children have received increasing attention

from medical services over the past several decades, and

this trend is continuing among several health related

professions, including clinical psychology.

Definition and Classification of Migraine

Migraine headache is defined by a list of symptoms.

Sacks (1992) notes that headache is never the sole feature

of a migraine nor is it considered to be essential. Several

requisite features have been identified, including the

presence of periodic or recurrent headache, nausea, and

family history. Other symptoms that may co-exist but that

are not essential for diagnosis include the presence of

neurological symptoms paresthesiaa, scotomata), or aura

(Bille, 1962; Rossi, 1989). Unilateral symptoms are common

and serve as a good marker for migraine in adults. These

will be discussed in further detail below.

The International Headache Society classifies headaches

into four broad categories, the first of which is migraine.

Subtypes of migraine in this classification scheme include

those without aura, with aura, ophthalmoplegic and retinal

migraine. The three remaining broad categories in this

system are tension-type headaches, cluster headaches, and

miscellaneous headaches not associated with structural

lesion. It should be noted, however, that this

classification system is designed for the adult population,

and has not been accepted for use in children.

Within the diagnosis of migraine, two common

classifications may be identified: classic and common


migraine. They are distinguished by the presence or absence

of aura. Rossi (1989) suggests that classic migraine, or

migraine with aura, is most likely on the same continuum as

common migraine, which is without aura. She proposes that

they differ more in the frequency rather than the quality of

the symptoms, and therefore should be considered as two

manifestations of one syndrome. Sacks (1992) identifies the

cardinal symptoms of common migraine to be headache and

nausea, while the classic migraine is characterized by aura.

Fenichel (1985) describes five broad groups of migraine

that differ slightly from those listed by the International

Headache Society. In addition to cluster headaches, and

classic and common migraine, they include basilar migraine

and migraine equivalents. Classic migraine may be

subdivided further into several categories, some which will

be mentioned briefly. Patients exhibiting focal motor

deficits, such as hemiplegia or ophthalmoplegia, can be

classified into a subgroup termed complicated migraine.

A common method of classifying migraine was provided by

Peroutka (1992). His list of subtypes and associated

symptoms is recreated in Table 1-1. Of interest is the

emphasis placed on the unilateral or contralateral

components of each migraine subtype. Again, however, this

system does not completely reflect migraine in children, as

they are more likely to exhibit bilateral than unilateral

headaches. A more important symptom in children is the

pulsatile nature of the headache. An interesting aspect

regarding the laterality of migraine is Wolff's (1963)

report that right-sided headaches are most common in right-


Table 1-1
Classification of Migraine

Types of Migraine

1. Common

2. Classic

3. Complicated

4. Ophthalmoplegic

5. Basilar Artery


Unilateral (80%)
throbbing headache with
associated nausea

Same as common migraine but
preceded by a visual aura

Sudden onset of hemiparesis
followed by a contralateral
throbbing headache

Unilateral eye pain and
ipsilateral ophthalmoplegia

Visual aura with alterations
in consciousness prior to

Source: S. J. Peroutka (1992). Migraine. in M. V.
Johnston, R. L. MacDonald, and A. B. Young (Ed.) Principles
of Drug Therapy in Neurology. Page 162.

These neurologic symptoms associated with migraine are

most commonly transient, resolving within 24 hours. Rarely,

though, they may leave permanent neurologic deficits that

can be quite subtle, diagnosed only with sensitive

neuropsychological test measures. Brown (1977) comments

that there is still doubt as to whether recurrent mild

ischemia can result in lasting brain-damage, with subsequent

dementia and cerebral atrophy. However, Cole and Aube

(1990) reported three well-documented cases of migraine that

were associated with intracerebral hemorrhage. Each of

these adult cases showed spasm of the external or internal

carotid artery. They concluded that "vasospasm associated

with severe migraine attacks may result in ischemia of the

intracranial vessel walls, leading to necrosis and

subsequent vessel rupture when perfusion pressure is

restored" (page 47).

Additional Symptoms

Although not specific to children, less frequent

symptoms may be present in a child diagnosed with migraines

in addition to those described above. While migraine is

often eased by sleep (Rossi, 1989), it may be worse in the

morning. This may raise the suspicion of increased

intracranial pressure effects. Rossi reports other, less

frequent symptoms that include paresthesia, a "march" of one

to thirty minute duration that progresses from upper limb to

face, and from distal to proximal. She also reports cases

of dysarthria or aphasia, or an acute confusional state

lasting anywhere from hours to days.

Additionally, in the acute phase of an attack, migraine

patients may show nystagmus, presumably due to

vertibrobasilar ischemia. Vertigo may also be a symptom

present in classical migraine or basilar migraine (Brown,

1977). Brown also reports that pectoralgia, or constricting

chest pain, may be an accompanying symptom of migraine.

The most common migraine symptoms associated with aura

are visual. These visual aberrations may include

experiences such as transitory blindness, sparkling lights,

micropsia, macropsia, or even visual hallucinations. Brown

(1977) describes many of these deviations that can range

from a simple blurring of the vision, to visual field cuts,

to complete transitory blindness. Colored lights in various

shapes may also be seen. Brown reports that hallucinations

of mice or butterflies, or an entire scene breaking from

reality, are not rare. Ophthalmoplegic pain and paralysis

of eye movement, while suspicious symptoms, may also

accompany migraine. Additional common symptoms are the

nausea and vomiting which often accompany a migraine attack.

This is sometimes referred to as "abdominal aura," and may

occur for a long period of time before the onset of actual

migraine symptoms (Brown, 1977).

Finally, another group of symptoms consist of brainstem

dysfunction, commonly referred to as the basilar artery

migraine syndrome. Patients with this subgroup of migraine

may exhibit ataxia, visual aura or even loss of vision,

vertigo, tinnitus, alternating hemiparesis, and parasthesias

of the fingers, toes, and corners of the mouth (Fenichel,

1985). Brief alteration in consciousness may also result.


Testing performed on children with basilar migraine has

shown an abnormal caloric response (nystagmus induced by the

introduction of cold or warm water to the ear canal)

(Eviatar, 1981). Others have shown an association with

epilepsy in this group, exemplified by spike wave forms on

EEG, recurrent seizures, and partial complex symptomatology

(Panayiotopoulos, 1980). Interestingly, antiepileptic

therapy yielded relief of migraine in this group of

patients. Fenichel (1985) asserts, however, that in the

absence of associated seizure activity, antiepileptics are

of no proven value in treating these children. As will be

seen, Fenichel's belief is challenged by current medical



If one is to group both classical and common migraine

together as elements of the same syndrome, then a singular

cause must be identified. Three prominent theories of

migraine pathogenesis will be presented here, as well as

other, more comprehensive, theories:

1. Vascular dysrequlation: A common hypothesis

proposed by Wolff (1963) and also held by Rossi (1989) is

that vascular dysregulation underlies both types of

migraine. Wolff (1963) conducted experiments and

observations that indicated a vasoconstrictive and

vasodilative phase, which distinguish the prodrome and

headache phase. Fenichel (1985) describes these same two

vascular phases of migraine. The first, which takes place

in the preheadache stage, consists of vasoconstriction,

during which cerebral blood flow decreases. This is most

strongly associated with the internal carotid system in

classic migraine. Symptoms associated with this phase

include the visual aberrations described above (or "migraine

equivalents"), which are the most common form of aura. The

second phase, which is associated with the headache, is

vasodilation, and involves the external carotid system.

Patients describe headaches during this phase as throbbing,

pulsating, or pounding, a characteristic of the vasodilative

phase. This second phase is also distinguished by the

concomitant symptoms of migraine such as nausea or vomiting.

Welch, Spira and Lance (in Diamond, Dalessio, Graham and

Medina, 1975) have shown that a single agent (such as

prostaglandin-e) can create these differential effects in

blood flow in external (carotid dilation) and internal

(redistribution of flow) vasculature.

While this theory of migraine has support from

research, others (Sacks, 1992) claim that it is far too

simplistic in and of itself to explain fully the

multitudinous aspects of migraine. The vascular components

of the migraine process may, therefore, be only another

aspect of migraine rather than an adequate explanation.

Other, more elaborate theories therefore exist in an attempt

to offer a better understanding of the entire process of


2. Serotonergic abnormalities: A second theory,

originated by Sicuteri (1961), arose from the finding that

plasma concentrations of serotonin decline at the appearance

of a migraine attack and remain significantly depressed

throughout the duration of the episode (e.g., Curran,

Hinterberger, and Lance, 1965; Hilton and Cummings, 1972, as

cited in Fenichel, 1985). In contrast, serotonin levels

rise during the migraine aura (Peroutka, 1992). This is

consistent with the evidence that serotonin applies a

powerful constrictor effect on extracranial vessels (Rigg,

1975; Sicuteri, Testi, and Anselmi, 1961). Current

hypotheses involving serotonin suggest that prophylactic

antimigraine medications function by blocking specific

serotonin receptor subtypes, whereas abortive drugs

stimulate these receptors. This may be too simplistic,

however, and, as will be discussed, the difference between

abortive and prophylactic medications may also be affected

by their action on different 5-HT receptor subtypes.

Further research has shown that the platelets of

migraineurs are different from those of nonmigraineurs in

their serotonin-releasing effect. Fenichel (1985) goes so

far as to say that migraine may primarily be a platelet

disorder, and daily aspirin may therefore act

prophylactically. The role of serotonin in migraine is

still unclear, but evidence supports at least partial

involvement of this neurotransmitter. Some authors (Sacks,

1992) report that one common symptom of migraine is

depression. This side effect is consistent with a serotonin

hypothesis given that serotonin is thought to play a strong

role in depression as well. Whatever the actual mechanism,

there is strong support for a serotonin hypothesis that is

based upon pharmacological evidence which has shown that

various medications that operate via serotonin can either

cause or terminate a headache episode (Diamond, Dalessio,

Graham, and Medina, 1975).

3. Spreading Depression of Leao: Finally, a

phenomenon known as the spreading depression of Leao has

been used to explain migraine pathogenesis. This electrical

phenomenon occurs in response to noxious stimulation

(Peroutka, 1992), and may be initiated by vasoconstriction

(Fenichel, 1985). This activity originates in the occipital

cortex, which may explain the predominantly visual aura.

The oligemia slows at the Rolandic and Sylvian fissures, but

may eventually reach the frontal cortex via the insula

(Fenichel, 1985). Its arrival at the sensorimotor cortex

occurs during the headache phase after the prodrome,

supporting the role of spreading depression in migraine.

In addition to the above hypotheses, the

trigeminovascular system also plays an important role in


migraine. Because it provides the primary afferent pathway

for head pain (Peroutka, 1992), this system has been

included in many theories of migraine pathogenesis. Neurons

in the peripheral unmyelinated fibers of the

trigeminovascular system contain substance P. This

neurotransmitter is known to be involved in several

functions: dilation of pial arteries, increase of vascular

permeability, and activation of cells involved in the

inflammatory response. This last function has an apparent

influence on head pain that can occur in the absence of

obvious external precipitating factors. According to

Peroutka (1992), the hypothesized action of the

trigeminovascular system and substance P may explain the

fact that the majority of migraines are unilateral. In

addition, the pain in migraine is referred to areas

innervated by the trigeminal nerve. Finally, the aura most

commonly associated with migraine is a result of "posterior

circulation dysfunction", this posterior area being densely

innervated by the first division of the trigeminal nerve.

Solomon's (1993) model of migraine incorporates several

concepts, some of which are common to the above mentioned

theories. First, he suggests that the trigeminovascular

system serves the purpose of protecting the brain, just as

other pain systems work to protect other parts of the body

by signaling problems, which, in this case, would include

ischemia or toxins. The trigeminal nerve can be stimulated

by both electrical (neuronal) and chemical factors. The

chemical factors would include neurotransmitters, such as

serotonin, and may work directly (hormones affecting

prostaglandins) or indirectly (e.g., REM sleep reducing

serotonin release from the dorsal raphe nucleus). This

latter mechanism may help explain why sleep often eliminates

the migraine in many patients. Whatever the trigger, the

stimulation may result in a release of tachykinins from the

trigeminal nerve, such as substance P, into the meningeal

and dural blood vessels. This results in the release of

histamine, vasodilation, and the platelet release of

serotonin. The release of substance P results in

inflammation and blood vessel swelling. Finally, this

swelling and inflammation causes distention of the cranial

arteries and the headache pain.

Solomon's model has the appeal of explaining many of

the components known about migraine, including a role of

serotonin, the cause of the actual headache pain, the

benefit of antihistamines, the purpose of sleep to alleviate

the migraine, and the possible mechanisms of numerous

triggers. The pain involved in migraine may be caused by

local inflammation and sensitization of nociceptors by

substance P. In any case, this model emphasizes the

importance of an ischemic or neurochemical noxious trigger

that stimulates the trigeminal nerve. This model does not


exclude other theoretical perspectives, but may instead help

to link them together in a more comprehensive model.

A more functional model of migraine was proposed by

Crisp, Levett, Davies, Rose, and Coltheart (1989). They

conducted a study on 57 migrainous adult patients, 26 of

whom had classic migraine with primarily unilateral

prodromata. Among these subjects, all were right handed,

and 19 had left hemisphere prodromata. One had right

hemisphere prodromata, but had a history of only one

migraine. Others were indeterminate or bilateral. The

findings indicated that unilateral classical migraine

categorized according to prodromal clinical characteristics

in right-handed people is exclusively a left (dominant)

hemisphere phenomenon. These researchers suggested that

this type of migraine was a result of a relative dysfunction

of the ipsilateral hemisphere, rendering it susceptible to

overload of information to be processed. In this case, left

hemispheric prodromata would be due to dysfunction in the

properties of the left hemisphere, such as verbal

processing. Although Crisp et al. did find some support for

their hypothesis, their interpretation may very well be

incorrect. That is, if an individual experiences a long

history of unilateral migraine (their subjects' mean length

of migraine history was nine years), then the processes

involved, such as those described by other theorists above,

may result in a compromise of the systems of that particular

hemisphere. Additionally, the stimuli that these

researchers used to measure spatial ability had qualities of

verbal processing and attentional demands rather than being

a relatively pure measure of spatial processing.

Nevertheless, the finding that right handers with unilateral

migraine have a strong tendency towards left hemisphere

involvement as measured by prodromata is intriguing. These

researchers screened out left handers for purposes of their

research, but reported that of the five subjects they found

with right hemisphere unilateral migraine, all five were

left handers. Although these researchers assessed numerous

patients with unilateral pain, persistent unilateral pain

may suggest other factors, such as arteriovenous


Demographic Variables


Although there are discrepancies, most place the

prevalence of childhood migraines at approximately four

percent (Bille, 1962). There is evidence that migraine has

a hereditary component, which is presumed to be autosomal

dominant, according to Fenichel (1985). Brown (1977) states

that as many as 90 percent of migraine cases have a family

history, also suggesting autosomal dominance. Fenichel

(1985) purports that the prevalence of migraine is 2.5

percent in children under 7, which then rises to 5 percent


in children past that age. In the older age group, females

are slightly over represented at a 3:2 ratio.


According to Fenichel (1985), the frequency and

intensity of migraine tends to decrease with age. However,

some studies have shown that, as a group, children with

migraine die at younger ages than nonmigraineurs (Leviton,

Malvea, and Graham, 1974). Also, presumably due to the

vascular mechanisms involved, this group has a higher risk

for stroke, especially in women taking oral contraceptives

(Fenichel, 1985).

Personality Differences

There have been several studies exploring personality

characteristics in children with migraines. Brown (1977)

asserts that, contrary to earlier beliefs, migraine is not

associated with high intelligence. Common findings from

more personality oriented investigations describe migraine

children as dejected, depressed, or withdrawn, with the

potential for becoming stubborn, inflexible, argumentative,

and rebellious (Wolff, 1963). Wolff also reported that

memory and attention can be impaired during the headache.

Bille (1962) has described these children as being more

anxious, fearful, and nervous than their same aged peers

without headaches. Most of the earliest studies employed

nonstandardized questionnaires or simply clinical impression

(e.g., Menkes, 1974). As a result, findings were mixed and


yielded little useful data. Others have attempted to avoid

such problems through the use of standardized instruments

and appropriate control groups.

Andrasik, Kabela, Quinn, Attanasio, Blanchard, and

Rosenblum (1988) conducted a study to examine the

personality characteristics of two groups of children with

migraines. An older, adolescent group consisted of subjects

between the age of 13 and 17, while the ages of children in

the younger group ranged from 8 to 12. All subjects were

age matched with controls. The measures utilized in this

study were primarily parent report and self-report

questionnaires. Findings from this study were that headache

sufferers earned significantly higher scores on measures of

depression from several scales, and expressed a higher

number of somatic complaints, even with the omission of

headache related questions. Andrasik et al. also found

that, overall, the adolescent group was less successful in

coping with their migraines than the younger group. They

hypothesized that adolescents, while trying to develop their

independence, at the same time had unpredictable migraine

attacks that often ruined social plans. This

unpredictability may therefore create a need to be dependent

on a caretaker or foster feelings of anxiety. Andrasik et

al. found that this pattern was particularly true for males,

and suggested that, because of accepted roles for males and

the expression of their pain or discomfort (which does not

include crying or openly admitting pain), male adolescents

had to find alternative methods of expressing their pain.

This often came in the form of externalizing behaviors or


In contrast to these findings, Cooper, Bawden,

Camfield, and Camfield (1987) found from their study of

children with migraine, that such children do not report a

higher number of life stressors than their best friends

without migraine. This was found despite the fact that the

children with headaches and their parents both reported that

stress is the most common trigger of migraine attacks.

Cooper et al. (1987) do agree with Andrasik et al. in the

finding by parent and self-report measures, that migraine

children have significantly higher ratings on scales such as

somatic concerns and depression. Andrasik et al. (1988)

suggest that migraine children do not experience greater

life stress (based on questionnaire) than normals, which

would suggest the differences in somatic concerns and

depression are related to the migraine rather than the

migraine resulting from stress. Furthermore, despite no

significant difference between the headache and control

groups on stress or anxiety, Cooper et al. (1987) did

provide evidence suggesting that the most anxious children

in their sample had the most frequent and severe headaches.

They concluded that stress and anxiety exacerbate migraines


in children who have inherited the migraine tendency, rather

than being a direct cause of migraine.

Neuropsycholoqical Differences

A recent paper by Kohler and Fennell (1992) examined

the differences in performance on neuropsychological tests

between adult migraine patients and normal controls. The

results of this study suggested that migraineurs perform

significantly worse on verbal and nonverbal memory tasks

than age matched controls. This difference was largely

attributed to the common migraine group, as there was no

significant difference between the classic migraineurs and

controls. No groups showed differences on learning tasks.

These researchers hypothesized that those migraineurs with

aura (i.e., classic migraine) may be able to cope better

with an impending migraine attack due to the benefit of a

warning sign, which would enable them to use medications

that may prevent the headache phase of the migraine.

Presumably, those patients who experience fewer headaches

may therefore be at a lower risk for ensuing ischemic damage

that could negatively impact memory functions.

Neuropsychological studies of patients with migraine,

however, have yielded mixed results. A study by

Leijdekkers, Passchier, Goudswaard, Menges, and Orlebeke

(1990) compared the neuropsychological performance of 37

adult female migraineurs with that of 34 nonheadache female

controls. They employed a variety of tests, including

measures of verbal learning, short term memory, reaction

time, sustained attention, and motor tapping. They did not

find significant differences on any of their measures of

cognitive functioning, nor did they reveal any effects of

migraine history or medication use. Further dividing the

migraine subjects into classical versus common also did not

result in group differences. They did, however, note higher

trait and state anxiety levels, state depression, and less

vigor, based on self-report measures, in the patient group

compared to the control group.

Yet another set of findings by Hooker and Raskin (1986)

provides evidence for a different interpretation of the

question of neuropsychological differences in migraineurs.

They administered a battery of tests including motor tests,

verbal and nonverbal memory tests from the Wechsler Memory

Scale, Trail Making Test, the Wisconsin Card Sorting Test,

and the Tactual Performance Test (TPT). Thirty-one migraine

subjects and 15 controls were used in this study, with a

combination of males and females. The authors found that

subjects with classic migraine exhibited decreased upper

extremity motor speed and dexterity, less efficient learning

of new associations between dissimilar symbols, and

dysphasic errors. Combining the classic and common migraine

subtypes into one group revealed poorer delayed free recall

for verbal material and decreased performance on the TPT.

These researchers argued that the classic migraineurs

demonstrated a larger scope of functional compromise than

the common migraineurs, and suggested therefore that the

classic form of migraine reflects a primary disturbance of

the central nervous system. Their findings are in direct

contrast to the study by Kohler and Fennell (1992), which

indicated that the common migraineurs experienced greater

neuropsychological impairment. Hooker and Raskin (1986)

also showed that no single class of drug showed differential

impact on the Halstead-Reitan Average Impairment Index,

which is consistent with the findings by Leijdekkers, et al.


In another study, Ai (1992) administered a

neuropsychological test battery to 48 migraineurs, 20

patients with functional headache, and 20 controls. The

primary result was that migraineurs had higher dysfunction

in the areas of motor, perception, memory, abstract thought,

attention, and information processing. Additionally, the

dysfunction was worse with a longer course of migraine.

Crisp, Levett, Davies, Rose, and Coltheart (1989)

demonstrated in their study of 57 male and female patients

that migraineurs as a group (i.e., classic and common

migraine) exhibited poorer verbal and spatial performance

than controls.

Despite the evidence resulting from these studies that

true differences do exist between migraineurs and matched

controls, another study provides evidence that there are no


significant differences. Burker, Hannay, and Halsey (1989)

examined 47 migrainous female college students and 24

controls. No significant differences were found on the

Halstead Reitan Neuropsychological Test Battery and other

memory tests. Like Leijdekkers et al. (1990), however, they

did observe personality differences as measured by the

Minnesota Multiphasic Personality Inventory (MMPI) on

Hysteria and Paranoia scales between classic and common

migraineurs, and differences on the Hypochondriasis and

Hysteria scales between all migraineurs and controls.

It is unclear what factors are contributing to these

divergent findings. One possibility is that many of the

findings by Hooker and Raskin (1986) are spurious, resulting

from the numerous statistical analyses they conducted on a

relatively small sample size. Kohler and Fennell (1992)

employed age and sex-standardized scores in a multivariate

analysis which limited the number of tests of significance

and controlled for the correlation between measures.

Leijdekkers et al. (1990) attempted to use relatively basic

behavioral tests rather than more complex tests, which they

felt would be overly sensitive to extraneous influences and

lead to spurious results. This may also have served to mask

real differences between the patient and control groups.

The diverse collection of research in this area does

not appear to support strongly any one hypothesis.

Different test batteries, methods of analysis, or individual

differences may all contribute to the inconsistencies

reported here. Additionally, these studies were conducted

on adult patients, which may be a vastly different

population than migrainous children, in part because of the

longer history of a potentially damaging process in the

adults. Also, children are less commonly diagnosed with

classic migraine. This dimension has been significant in at

least several studies and may lead to different findings for

an adolescent population. Unfortunately, no comparable

research base exists to date on children with migraine.

Imaging Techniques and Coincidence with Epilepsy

Some investigators assert that the use of EEG, cranial

computed tomography (CT), or other types of brain imaging

techniques, constitutes an excess in diagnostic procedure

with migrainous children (Prensky and Sommer, 1979).

Others, however, declare that such diagnostic tools have

value in the research and assessment of migraine. For

example, Alvarez-Cermeno, Gobernado, Freije, Zaragoza, and

Gimeno (1984), reported two cases of pediatric migraine

which showed CT contrast enhancement in the occipital

region. They stated that although these areas were poorly

defined, the important finding was that the CT changes were

evident only during the active phase of migraine.

Regional Cerebral Blood Flow (rCBF) has also been

examined in children with migraine. Roach and Stump (1989)

concluded that regional reduction in the expected blood flow

surge after CO2 inhalation was a frequent finding, lending

support to the already strong belief that pediatric migraine

is a disorder of vasomotor function. However, others argue

that there may be no change in cerebral blood flow in

migraine without aura (Silberstein, 1991). In a study on

adults with migraine, Thie, Carvajal-Lizano, Schlichting,

Spitzer, and Kunze (1992) attempted to measure cerebral

vasoreactivity (CVR) via trascranial Doppler ultrasound

(TCD) during cognitive and motor tasks. They imaged the

left middle cerebral artery (MCA) and left posterior

cerebral artery (PCA) and compared CVR in migraineurs to

controls. Although they found that migraineurs as a group

had higher mean flow velocities for photic stimulation,

observation of complex images (both PCA measures), and the

cognitive task (MCA measure), they cautioned against any

strong conclusions. They stated that individual differences

were too great to allow this form of imaging to be a

sufficiently sensitive measure of migrainous activity.

Migraine is not an uncommon associate of epilepsy,

particularly when the migraine is associated with an aura

(De Romanis, Buzzi, Feliciani, Assenza, and Agnoli, 1991).

Fenichel (1985) asserts that familial epilepsy syndromes and

migraine are associated for the following reasons:

(1) they are both familial, paroxysmal, and associated
with transitory neurologic disturbances;
(2) there is an increased incidence of epilepsy in
migraineurs and migraine in epileptics;
(3) headache can be a seizure manifestation; and

(4) abnormal electroencephalograms are common in both
disorders. (p. 80)

The mere presence of a correlation, however, does not

confirm that epilepsy and migraine are biologically linked.

Furthermore, abnormal EEGs are not indicative of any

specific disorder, as some portion of the normal population

shows evidence of abnormal electroencephalographic activity.

Other similarities exist, however. It may be that

repeated migraine creates compromised tissue that could

result in epileptic discharges. Also, when a child has

symptoms of confusion, personality change, rage reactions,

and hallucinations, they may mimic the symptoms of a partial

complex seizure. Furthermore, the flicker-rate of scotomata

in migraine is of the same frequency of the stroboscopic

illumination that is most likely to cause epileptic seizure

(Sacks, 1992). Jackson (1931, as cited in Sacks, 1992)

asserted that migraines were in fact cases of epilepsy of a

sensory nature, with the headache and vomiting representing

the post-paroxysmal period. Other evidence does lend

support to the correlation between epilepsy and migraine,

although the precise relationship is unknown.

EEG recordings between migraine attacks may reveal

sharp or spike waves from the temporal-occipital or

temporal-parietal regions, and reflect clinically reported

symptoms from children with migraine. These children often

report visual hallucinations, hemianopsia, or amaurosis in


conjunction with migraine attacks (De Romanis et al., 1991),

which is consistent with photophobia reported in migraine

sufferers without epilepsy. In many cases, migraine is seen

in combination with epilepsy such that an attack may be part

of the prodromal phase of the epileptic event, but there may

also be a dissociation between migraine and epilepsy such

that the migraines continue years following the resolution

of the seizure disorder.

Because the occipital EEG wave pattern is so common in

childhood migraine headaches (e.g., De Romanis et al.,

1991), it may be that even in the absence of clinically

evident seizure behavior, subclinical abnormal activity is

responsible for migraines. Why the posterior cortical areas

are more susceptible to such activity may be due to a

neuronal sensitivity in the occipital region (De Romanis et

al., 1991). This, in conjunction with the fact that visual

patterns may trigger migraine as well as epilepsy (Brown,

1977), lends support to a relationship between the two

syndromes. In fact, Sacks (1992) considers migraine to be

on the "borderland" with epilepsy and other neurological

disorders, suggesting that they share similar symptomatology

and possibly similar underlying mechanisms. He suggests

that the paroxysm in migraine can be 20 times slower than in

the epileptic counterpart.

Additional evidence from EEG data on migrainous

children reveals no difference between these children and

age-matched controls between headaches. However, Seri,

Cerquiglini, and Guidetti (1993) found that during visual

aura a decrease in occipital alpha power contralateral to

the affected hemifield was evident in all migraine patients.

Furthermore, this was followed by an increase in delta power

in the bilateral frontal regions. During the headache, an

increased delta activity in posterior-temporal and occipital

electrode sites was also shown. The involvement of

occipital areas, especially during visual aura, is

consistent with neuroanatomy. The bilateral frontal

involvement, however, may reflect neuropsychological

impairment, probably quite subtle, that may include problems

with attention. Whether such an impairment could be

measured between events remains to be seen. Such

differences in EEG background activity (slowing) has been

observed between migraineurs, idiopathic epilepsy patients,

and controls, with no significant difference within the two

patient groups (Farkas, Kohlheb, Benninger, and Matthis,


Like migraine, epilepsy may manifest with the

predominant associated feature of abdominal pain, sometimes

termed "abdominal epilepsy". The differentiation between

abdominal migraine and abdominal epilepsy is not entirely

clear cut on the basis of subsequent convulsions or EEG

(Brown, 1977). Prensky (1976) reports that children with

abdominal pain as their major symptom seem to have more

epileptiform EEG changes that those with simple migraines.



The most common factor implicated in triggering a

migraine is psychological stress (Rossi, 1989) or emotional

tension (Brown, 1977). In pediatric cases, the frequency of

migraine is much higher during the school year than in the

summer, also implicating psychological stressors. This

finding suggests that an ideal treatment for migraines is


Behavioral approaches to the treatment of pediatric

migraine consist primarily of biofeedback, which can be

defined simply as bringing those bodily processes that are

normally involuntary and unconscious under voluntary control

(Gascon, 1984). This often requires the use of electronics

that have the capability of translating biological signals

(e.g., vasodilation) into signals that can easily be

detected (e.g., auditory signal). One method of biofeedback

is EMG, which is commonly used to assess frontalis muscle

activity, typically in tension headaches. For migraine,

however, temperature biofeedback is commonly employed as a

behavioral technique designed to influence vasoconstrictive

processes. While many of the treatment studies conducted in

adults have not yielded strong findings (Gascon, 1984),

children may be more promising candidates for this type of

treatment. Womack, Smith and Chen (1988) report that

children naturally like the use of imagery and biofeedback,

and are often enthusiastic about the procedures. Gascon

(1984) has suggested that children learn biofeedback

techniques more quickly than adults, and that if effective,

such a treatment option has obvious desirability. Brown

(1977) recommends that regular drug therapy be reserved for

only the most severe cases. While medications are still

most common, behavioral approaches offer treatment that

allow the child to feel more in control of their migraines,

as well as to avoid the problems of drug side effects.

In a study conducted by Womack, et al., (1988),

children were asked to keep diaries of their headaches.

They then participated in biofeedback (EMG and/or skin

temperature for most subjects), as well as relaxation and

mental imagery. After eight weekly sessions and 12 month

follow-up, results showed that, of 29 subjects, 48 percent

of the migrainous patients were headache-free, and an

additional 41 percent showed a greater than 50 percent

reduction in headache frequency. Eighty-six percent showed

a decrease of over 50 percent in perceived intensity.

Despite the fact that Womack et al. did not employ a control

group, these findings are encouraging. Many factors may

have been involved, however, such as the simple act of

monitoring the pattern of headaches through the use of a

diary, or even a treatment placebo effect, and the

investigators made no attempt to tease out possible

differential effects. Also, since migraine headaches may be

erratic in their pattern of occurrence, some changes may

have been attributable to normal variations in occurrence.

Additionally, the short duration of the study may not be

sufficient to determine if there was an actual change or a

temporary placebo effect. Also, each subject received a

combination of treatments (imagery, relaxation, and

biofeedback). This would make it difficult to determine

what aspects of treatment, if any, played a significant

role. Despite the shortcomings of the study, the authors

conclude that behavioral treatment has a high probability of

improving headache condition in children and adolescents.

Guarnieri and Blanchard (1990) attempted to assess the

effectiveness of biofeedback in a clinic treated group

versus a home practice group. They claim to have found a

clinically significant reduction of headache frequency and

intensity, though these numbers did not reach statistical

significance. There was no significant difference between

the two groups on measures of expectation or symptom

improvement, suggesting that home treatment is equally

effective. Burke and Andrasik (1989) found similar results,

adding that the gains made by their subjects were stable,

maintained through a one-year follow-up period. Finally,

Smith, Womack, and Chen (1989) found no correlation between

age, sex, headache type, hypnotizability, and clinical


outcome. Most recently, these findings were strengthened by

the similar results of McGrath, Humphreys, Keene, Goodman,

Lascelles, Cunningham, and Firestone (1992) who showed that

self-administered behavioral treatment was effective with

adolescents, even at a one year follow-up. An adult follow-

up study on the long term effects of training in relaxation

and stress coping was conducted by Sorbi, Tellegen, and Du

Long (1988). They found that the positive results from

behavioral and cognitive treatments were still in effect

after a period of three years.


Some researchers have considered allergies and diet as

possible triggers for migraine attacks (Brown, 1977),

including foods rich in tyramine or phenylethylamine.

Consequently, a modified diet has been attempted in the

treatment of migraine. Specifically, tyramine, which is

known to release serotonin from platelets, has been reduced

in diets for clinical treatment, or added to the diets of

controls to precipitate migraine experimentally. This

relationship has yet to be firmly established.


Some common medications used in the treatment of

migraine headaches in children include propranolol,

ergotamine tartrate, pizotifen, cyproheptadine, and calcium

channel-blocking agents (Rossi, 1989). Some of these (e.g.,

ergot derivative) are most effective when administered

during the aura, if present, prior to the onset of the

actual migraine. Fenichel (1985) suggests the use of two

medications, ergotamine and isometheptene mucate, for the

acute treatment of migraine. For ergotamine, the chief

mechanism of operation is constriction of the extracranial

blood vessels. The latter drug, isometheptene mucate or

Midrin, is a sympathomimetic agent which constricts

extracranial veins, and has been shown by a few studies to

be at least as effective, if not more, than ergotamine.

Brown (1977) suggests that ergotamine will ease headache of

common migraine, but may actually aggravate the ischemic

neurological abnormalities of complicated migraines.

For migraine prophylaxis, Fenichel (1985) lists many

drug classes, including serotonin agonists and antagonists,

antidepressants, antihistamines, vasoconstrictors,

vasodilators, tranquilizers, calcium-blocking agents, and

antiepileptics. Solomon (1993) states that prophylactic

medication operating via serotonin are specific to 5-HT2

receptors. Antagonists "are able to inhibit serotonin from

inducing an arachidonic acid-derived inflammatory state."

(p. 202). This explains why such a medication can operate

only prophylactically, because once the inflammation has

been initiated, the 5-HT2 antagonist would be useless.

Gascon (1984) also lists antiepileptic drugs as being

among the group of effective medications for migraine, and

asserts that these drugs are most effective in preadolescent

children. Phenobarbital, phenytoin, carbamazepine, and

valproic acid are among the many antiepileptic drugs used

for this purpose. Moreover, there is no clear indication

that associated epileptiform activity is necessary for the

effective use of antiepileptics in children with migraine.

The mechanism in which antiepileptic drugs affect migraine

is yet to be discovered, and to date, there is a surprising

lack of well conducted studies (e.g., prospective, double

blind, placebo controlled) in this area. However, according

to Solomon (1993), the action of these medications, at least

that of valproate, would operate at the level of 5-HT2

receptor site in order either to prevent serotonin-induced

cerebral vasospasm and ischemia, which may activate the

trigeminal nerve, or by inhibiting the arachidonic acid

metabolism as described above.

Sorensen (1988) conducted a prospective study of the

efficacy of valproate in 22 adult patients with severe

headache resistent to previous prophylactic treatment.

Subjects were followed over an average of six and a half

months to assess the efficacy of the medication. Eleven

patients were free from migraine, 6 demonstrated a

significant reduction in frequency, and one had no change.

Four of the patients withdrew from the study because they

could not tolerate the withdrawal of their previous

medication (injected narcotic). Sorensen also attempted a

withdrawal of the valproate for four weeks with three

patients who were free from the attacks. In all three

patients, their headaches recurred, but subsequent

administration of valproate again caused the disappearance

of the headaches. Side effects noted included drowsiness in

one patient and slight weight gain in three females, and two

patients reported severe cold painful paresthesias of the

extremities only during the first days of treatment.

Unfortunately, a placebo effect cannot be ruled out in this

study because no control group was used and patients and

experimenters were unblinded. However, because these

patients had previously tried numerous medications without

apparent relief, it is not likely that these results are

solely due to a placebo effect.

Cognitive Effects of Anticonvulsant Medication

Despite the number of studies conducted for the purpose

of examining the differential cognitive or

neuropsychological effects of antiepileptic medication, very

few have employed good experimental methods such as double-

blind designs, adequate control groups, or appropriate

neuropsychological tests. Therefore, a need remains for

solid research to determine the extent to which various

antiepileptics affect cognitive and behavioral aspects of

children with or without epilepsy. In contrast, somewhat

more information exists on the effects of beta blockers and

analgesics, and suggests that both of these medications

negatively affect verbal memory and learning (e.g., Kohler

and Fennell, 1992), at least in adults.

Some authors (e.g., Meador, Loring, Huh, Gallagher, and

King, 1990) contend that most of the more popular

antiepileptics have similar treatment efficacy, and

therefore differential cognitive effects should be given

appropriate consideration. Trimble and Cull (1988) reviewed

over a dozen studies on the effects of several

antiepileptics in children with varying types of epilepsy

and summarized their findings as follows: Phenobarbital

substantially impairs behavior, whereas valproate and

carbamazepine have little effect. Other medications have

far less data and await further research.

Adult Studies of Cognitive Effects of Valproate

In a longitudinal study of adults with a variety of

types of epilepsy, Dodrill and Wilensky (1992) followed 198

patients for five years that were on phenytoin monotherapy

or polytherapy. They concluded that there was no evidence

that long term administration of phenytoin is associated

with loss of cognitive abilities over time. Instead, they

infer that it is the effects of an epileptic condition that

cause cognitive decline, and because their subjects had

milder cases of epilepsy, no deterioration could be seen.

In order to lend support to their belief that it is the

epilepsy that causes the cognitive changes and not the

medication, drug trials on non-epileptic groups would be


beneficial. Because their study included adults only, they

admit that their results should not be applied to children

without further research on younger, pediatric populations.

Corbett, Trimble, and Nichol (1985) did show cognitive

deterioration in children on long-term antiepileptic

therapy, with those with higher phenytoin levels having

lower performance IQs. Verbal IQ was not affected.

Gallasi, Morreale, Lorusso, Procaccianti, Lugaresi, and

Baruzzi (1990) employed a withdrawal of medication design in

an adult epilepsy group of 20 patients who were seizure-free

for at least two years. They tested their subjects at four

points in time, ranging from baseline to 21 months. They

found significant differences between the epilepsy group and

controls on measures of reaction time and a global

performance measure that included all tests. This

difference was found at baseline and the second testing

(drug reduced to half). They compared the results to normal

controls who were only tested once, and asserted that the

gradual improvement they found was due to the gradual

removal of valproate. However, a significant practice

effect could account for this finding, since the controls

were not tested repeatedly. It is also noteworthy that

their findings, although statistically significant, were not

clinically significant, as none of the patients had noticed

any changes in the abilities measured. This finding is

consistent with the literature, which has shown only mild

differences, if any, as a result of valproate.

A study conducted by Sommerbeck, Theilgaard, Rasmussen,

et al. (1977) evaluated the psychotropic effects of

valproate in a heterogenous sample of epileptics. The

patients were given either valproate or placebo in

conjunction with other medications. This is a drawback for

this study that purported to find minor differences with

concurrent administration of valproate, because the impact

of the other medications cannot be assessed adequately.

Measures used included Digit Span, Paired-associate

learning, simple reaction time, Stroop's Color Naming test,

Hidden Patterns test, time estimation, a cancellation task,

a visual gestalt test, and motor tapping. The results

suggested that valproate caused declines in psychomotor

tempo and visuo-spatial analytic and synthetic functions.

However, these results are also suspect due to the small

sample size (20 patients) and the heterogenous nature of the

sample (e.g., the ages ranged from 13 to 63 years). Such

extraneous factors do not allow for strong conclusions.

Bittencourt, et al. (1992) also examined the three

medication groups (valproate, phenytoin, and carbamazepine)

in adults during drug administration. They found that the

epilepsy patients on phenytoin performed significantly worse

on the immediate recall for pictures than controls, those on

carbamazepine performed worse on the Stroop test than

controls, and those on valproate did not demonstrate any

appreciable difference from controls on these tests.

In addition to the research involving patients with

epilepsy, some researchers have attempted to assess the

effects of antiepileptics in normal controls. For example,

Boxer, Herzberg, and Scott (1976) found that a single dose

of valproate (400 mg) had hypnotic effects on medical

students used as controls. While this effect was greater

with the concurrent administration of phenobarbital, the

important finding was that valproate alone could produce

such effects. This study, however, was not without

methodological problems. The researchers only used eight

male subjects who were given only three psychological tests.

These tests consisted of a subjective report of sleepiness,

Digit Symbol, and a card sorting test (the authors did not

specify which). None of these measures showed any

differences. Their findings were based exclusively on EEG

record during a 30 minute period after only one dose of the

medication, one hour after administration. Perhaps the

biggest drawback to the study was their failure to measure

blood levels of the drugs used. Although the dosage was

equal for each subject (400 mg of valproate), there was no

apparent attempt to assess if the drug was in a therapeutic

range that would make the comparison more relevant to

clinical use.


In a better designed study, Thompson and Trimble (1981)

studied the effects of sodium valproate on cognitive

functioning in normal volunteers. Subjects were

administered 200 mg of sodium valproate or placebo for a

period of two weeks in a double-blind cross-over design. No

significant effects on memory, concentration, perceptual

speed, or motor speed were found. However, subjects did

require a significantly longer amount of time to answer

questions in a decision making task. Several reasons for

these results are possible. The dosage was less than that

used in the Boxer, Herzberg, and Scott (1976) study and may

therefore not create as strong an effect. The tests used by

Thompson and Trimble (1981) are not particularly demanding,

as memory was assessed through a simple recognition

paradigm, and other tasks were well within the range of

abilities for such a normal subject group (i.e., medical

students). Therefore, assessing the effects of valproate

would seem to necessitate using sensitive test measures to

reveal differences between control and experimental groups.

In order to investigate the effects of valproate on

sleep, reaction times, and visual evoked potential (VEP),

Harding, Alford, and Powell (1985) administered placebo,

low-dose, and high-dose conditions to 10 normal adult

volunteers. Withdrawal of medication effects were also

assessed. They found no significant treatment effect for

simple reaction time, nor did they find any differences in

either the latency or amplitude of the VEP. A decrease in

rapid eye movement and an increase in delta activity during

sleep was seen under the high-dose condition. Their lack of

findings could be a result of their small sample size,

particularly for the reaction time data, which was available

for only 6 of the 10 subjects due to technical difficulties.

The results from this study are consistent with those of

Thompson and Trimble (1981), however.

Child Studies

Duncan, Shorvon, and Trimble (1990) attempted to

examine the differences between several antiepileptic agents

(i.e., phenytoin, valproate, and carbamazepine) when they

were withdrawn from 58 children with partial and generalized

epilepsy. This method of assessing improvements following

the discontinuation of medication has the drawback of

assuming that these drugs have no residual, long term

effects. In fact, Duncan, Shorvon, and Trimble asserted

that their results supported the view that, since removal of

valproate resulted in the least improvement on

neuropsychological tests, it had the least negative impact

on the children who were taking it. This conclusion, while

it may have some validity, could be a misinterpretation of

an opposite effect. That is, if valproate has the more

detrimental, long lasting cognitive effects of the three

drugs, the results would have been the same. Regardless of

the final interpretation, Duncan, Shorvon, and Trimble found

that finger tapping was improved upon removal of all three

drugs, while letter cancellation, digit symbol substitution,

digit span, and serial subtraction, were not significantly

affected by the removal of valproate.

In a similar design, Blennow, Heijbel, Sandstedt, and

Tonnby (1990) reported tentative results in the midst of

their ongoing study. At the time they reported these

interim findings, 69 children with various types of epilepsy

had been studied, along with 69 matched controls (the

authors did not specify how control subjects were matched).

They described patterns such as faster reaction time at the

expense of accuracy for the valproate and carbamazepine

groups, slightly decreased motor fluency in the right hand

for the valproate and phenytoin groups (in contrast to

Duncan et al., 1990), and short-term memory decreases in all

groups treated for epilepsy. The last finding is difficult

to separate from the effects of long term epilepsy, however,

despite their use of a control group. The heterogenous

epilepsy group is also a common drawback in the study of

cognitive effects of antiepileptic medications, particularly

when the different types are not considered in analysis.

Vining, Mellits, Dorsen, et al. (1987) conducted a

double-blind counter balanced cross-over study of 21

children with epilepsy, comparing the cognitive impact of

phenobarbital and valproate. Patients were given a

neuropsychological battery three times: at baseline, after

six months on one medication, and after six months taking

the other medication. The battery consisted of tests

including intellectual assessment (Wechsler Intelligence

Scale for Children-Revised), measures of attention span and

sustained attention, motor tasks, achievement tests, mazes,

and the Bender-Gestalt, among others. While both

medications controlled seizures equally well, these authors

found that those patients on the phenobarbital performed

worse on certain subtests of the WISC-R (including verbal

and performance subtests), paired associate learning,

reaction time for errors on a continuous performance test,

and arithmetic achievement scores. Further, they found that

the patients on valproate earned more favorable scores on

parent questionnaires than those on phenobarbital. While

intriguing, these results must be considered tentative

because of the statistics used. With only 21 subjects,

these authors employed over 90 t-tests. Their findings may

be spurious rather than reflecting actual differences

between the side effects of the two drugs.

To assess the effects of dosage, fluctuations in

concentration, and diagnosis, Aman, Werry, Paxton, and

Turbott (1987) studied the effects of valproate in 46

children with epilepsy who received valproate monotherapy.

Three sets of tests were administered: measures of short-

term memory, continuous performance task, mazes, pursuit

rotor task, and Matching Familiar Figures, with one week

intervals. The timing was altered to take advantage of

medication dosing (one test was administered pre-dose,

another post-dose), with the first test administered to

minimize subsequent learning effects. Subjects with lower

doses performed better on the psychomotor tasks than those

with higher doses. Timing of the dosing (pre- or post) did

not have an effect. Again, these authors over used the t-

test without considering their sample size. The age-range

of their patients was also rather broad (4.4 years to 15.4

years), making accurate interpretation of their data


Forsythe, Butler, Berg, and McGuire (1991) utilized

more rigorous criteria for both subject population and test

measures. They targeted only new cases of epilepsy with no

neurological deficit, who had had either three tonic-clonic

seizures, three complex partial seizures, or three partial

seizures with secondary generalization. They employed an

appropriate experimental design, and monotherapy as opposed

to polytherapy. Michael Trimble served as consultant on

choosing appropriate test measures that would be most

sensitive, least affected by practice effects, high in

convenience, and would cause minimal fatigue and boredom.

Subjects were randomly assigned to one of three medications:

carbamazepine, phenytoin, and sodium valproate. After six

months of treatment, children on sodium valproate performed

significantly better on memory scores than children on


carbamazepine, and this difference was even greater after a

year. Speed of information processing was impaired in both

the carbamazepine and phenobarbital groups, this being

evident after only one month of treatment. The better

performance on memory scores with those on valproate

relative to those on carbamazepine may represent an actual

drug effect with the carbamazepine exerting a more negative

influence on memory. However, much of the research in this

area has employed tests that may not be as sensitive to

subtle, sub-clinical effects of memory and attention.

Future research should target these areas with measures less

susceptible to a ceiling effect or more likely to reflect

subtle differences.

In summary, several broad conclusions can be made about

the cognitive effects of antiepileptic medications.

Consistent with Vining's (1987) descriptions, these drugs

can be ranked by the degree to which they affect cognition.

Phenobarbital, which historically has the worst effects on

cognitive functioning, has its primary effect on short-term

memory, but also adversely affects long-term memory,

concentration, and symbol manipulation. Phenytoin

selectively impairs abilities related to problem solving and

visuomotor tasks, timed motor task, and attention.

Carbamazepine appears to have similar effects, although at

lower magnitudes. Finally, valproate appears to have

minimal effects, but may influence the most subtle aspects


of cognition. This makes sensitive neuropsychological tests

requisite in the assessment of differential effects.

Additionally, the research on the cognitive effects of

valproate is extremely mixed due to differences in

assessment tools, different populations, and varied

methodology (i.e., patients on polytherapy, blind versus

unblind, the use or lack of adequate controls). While some

research has been conducted on normal adult populations in

this area, comparable studies on children are lacking, in

part due to ethical considerations. Research on a non-

epileptic pediatric group may provide the necessary data to

assess the effects of the medications in children.

Characteristics Specific to Valproate

Method of Action

The absolute method of action of valproate is

uncertain, although GABA has been the hypothesized

neurotransmitter involved. Typically given orally,

valproate is rapidly absorbed and reaches peak serum levels

approximately one to four hours after a single dose. The

serum half-life is typically 6 to 16 hours. Meals do affect

the rate of absorption, but only slightly.

While some support for GABA's involvement in valproate

exists, Johnston (1984) cites contradictory evidence. His

position on the GABA mechanism is that changes caused by

valproate cannot be attributed to an increase in the GABA

level, the potentiation of postsynaptic GABA responses, or

any direct membrane effect on neurons. Such an assertion

would appear to refute GABA's involvement. However, the

research on the mechanisms of valproate are far from

resolved, and the GABA hypotheses are not refuted as yet.

Animal models have shown that valproate can increase

GABA by inhibiting GABA transaminase. The significance of

this, however, is questionable, because large

administrations of valproate (as much as 20 times a clinical

level) results in only modest and transient increases in

GABA levels (Emson, 1976). Valproate is, however,

considered to be a GABA agonist, which may affect migraine

via its influence on cerebral arteries and circadian rhythms

(Solomon, 1993). GABA dilates the cerebral arteries and may

help regulate hormone secretion from the anterior pituitary


It has also been purported that valproate has an effect

on serotonin metabolism, which would make intuitive sense

based on the potential role of serotonin in both migraine

and epilepsy (Van Woert and Hwang, 1978). In an animal

study of the effect of valproate on serotonin metabolism,

Hwang and Van Woert (1979) found that a single injection of

valproate (400 mg) decreased total serum tryptophan and

increased free serum tryptophan, brain tryptophan, and brain

5-hydroxyindoleacetic acid (5HIAA). However, they found no

effect on brain serotonin levels. They concluded that

valproate may act by displacing protein-bound serum

tryptophan which would result in an increase in brain

tryptophan, which would then be converted to serotonin.

In addition to proposing a mechanism for valproate and

serotonin (5-HT), Hwang and Van Woert also found differences

for various brain regions. For example, the largest effect

was a 46 percent increase in 5HIAA in the hippocampus.

Other areas showed increases as well, including the

hypothalamus, medulla, cortex, and striatum, although these

changes were not as large as that found in the hippocampus.

Despite the fact that these researchers did not report

differences in whole brain levels of 5-HT, there was a

significant increase found specifically in the striatum.

This level was elevated 113 percent over that of control

rats for this region. The overall elevated concentrations

of the various assays were attributed to increased 5-HT

synthesis. The authors suggested that the synthesis may

also be enhanced by 5-HT uptake blockade.

Using an animal model similar to that of Hwang and Van

Woert, MacMillan (1979) failed to support the hypothesis

that valproate acts to increase 5-HT synthesis. He argued

that the increase in 5HIAA was from decreased transport of

the metabolite out of the brain. An important difference

between his study and that by Hwang and Van Woert (1979) is

the lack of exploration into specific brain regions in the

MacMillan study. He acknowledged this as a potential

drawback, and it may hold the answer to the different

conclusions made from each study. Hwang and Van Woert

suggested that valproate may affect specific neuronal pools

(such as the striatum) that would obscure results from

studies that examined entire cortical levels of 5-HT, such

as in the MacMillan study.

Physiological Side Effects

As discussed above, valproate does not appear to have

as negative an effect on cognition as other antiepileptic

medications. However, similar to any drug, it does carry

certain risks to physiological functions. According to the

Physicians Desk Reference (PDR), valproate's most

significant physical side effect is hepatic failure.

Although not a common effect with valproate alone, signs of

sedation can be seen more commonly in conjunction with other

medications, particularly other anticonvulsants. Tremor is

a dose related side effect. More rarely, additional CNS

effects include ataxia, headache, nystagmus, diplopia, and

incoordination. It should be noted that none of these

symptoms are particularly common, and therefore do not pose

a significant threat to the child, although careful

monitoring is essential. Mattew and Ali (1991) noted that

some of the possible side effects may evidence themselves

frequently in certain subjects, such as hair loss (which

actually represents a speeding up of the normal process of

shedding of old hair), tremor, and weight gain.

Carnitine as a Supplement to Valproate Therapy


Research on the physiological effects of valproate have

revealed that children on such medication have significantly

lower plasma and erythrocyte free carnitine concentrations

than children on alternative antiepileptic medications

(Thom, Carter, Cole, and Stevenson, 1991). Carnitine is an

important nutrient found in the diet, plays an essential

role in the oxidation of fatty acids and in ketogenesis, and

some researchers have correlated a deficiency of this

nutrient with hepatotoxicity. Potentially toxic acyl

compounds may accumulate when the oxidation process of these

elements is impaired. Carnitine functions to regulate the

ratio of free coenzyme A (CoA) to acylcoenzyme A (acyl CoA)

in the mitochondrion (Coulter, 1991). The results from one

study indicated that free and total plasma concentrations of

carnitine decreased significantly after 15 days of valproate

administration in epileptic patients as compared to normal

controls (Riva, Albani, Gobbi, Santucci, and Baruzzi, 1993).

Murakami, Sugimoto, Nishida, Woo, Araki, and Kobayashi

(1992) showed in their research with rats that this change

in carnitine level resulting from valproate administration

could be altered to a more normal state with the concurrent

administration of carnitine. Giovannini, Agostoni, and

Salari (1991) argue that carnitine is an essential nutrient

for children, affecting multiple physiological functions

essential for development and growth. Therefore, it is


possible that carnitine may be supplemented for children who

are prescribed valproate to counteract such a deficiency.

Carnitine may not completely counteract the negative

effects of valproate, however. Siemes, Naue, Seidel, and

Gramm (1992) reported a case study of an infant who died

secondary to valproate complications despite the adjunctive

administration of carnitine. This finding was supported by

the research of Thurston and Hauhart (1992) who found that

carnitine administered to rats who had been given valproate

showed no effect on the reduction of free CoA and acetyl CoA

levels in the level. They found that this effect was only

achieved when carnitine was administered with pantothenic

acid. They asserted that their findings support the

hepatotoxic theory that valproate depletes CoA via

inhibition of B-oxidation and associated enzymic reactions.

Because CoA or acetyl CoA act as necessary substrates or

activators for approximately 100 synthetic and catabolic

enzyme reactions, the diverse effects of valproate can be

explained through this theory. The CoA levels are

diminished by the action of valproate, which may sequester

the substance into forms that are poorly metabolized. In

children with epilepsy, Melegh, Kerner, Acsadi, Lakatos, and

Sandor (1990) demonstrated that administration of carnitine

in conjunction with valproate over a period of 14 days did

not change the low plasma level of B-hydroxybutyrate. Total

carnitine levels were found to be lower in the children with

epilepsy compared to normal controls: however, this

difference was eliminated with the administration of

carnitine. Hypoketonemia was found to be carnitine

independent, though, which leaves the question of the

hepatotoxic role of valproate and the administration of

carnitine unanswered.

While research on the physiological impact of carnitine

deficiency is limited, cognitive or psychological effects of

such a deficiency in the face of valproate administration is

nonexistent. A need clearly exists for such research. One

potential reason for this lack of research is that there are

two forms of carnitine: L-carnitine and acetyl-L-carnitine.

The former is used in conjunction with valproate in some

cases to help prevent hepatotoxicity. Acetyl-L-carnitine, a

supposed cognitive enhancer, has been used in a few studies

of dementia and cognition, with its influence on the

facilitated production of acetylcholine, particularly in the

hypothalamus. Sinforiani, lannuccelli, Mauri, Costa, Merlo,

and Bono (1990) examined the neuropsychological effects of

this form of carnitine supplement in a population of

demented patients. They found that, compared to

administration of piracetam, a neurotropic purported to

improve learning and memory, patients given acetyl-L-

carnitine showed significant improvements in attention,

comprehension ability, and specific behavioral aspects.

They also asserted that carnitine has been found to improve


memory impairment. Although the two forms of carnitine are

related (acetyl-CoA + carnitine acetylcarnitine + CoA),

they are typically used for different purposes, prohibiting

comparisons of their cognitive effects. However, Thom,

Carter, Cole, and Stevenson (1991) suggest that signs of

encephalopathy, in addition to confusion, may be a

neurologic symptom associated with severe L-carnitine


Studies of Valproate in the Treatment of Migraine

While valproate has not become a drug of choice for the

treatment of migraine, there is some support for its

efficacy. For example, a study by Hering and Kuritzky

(1992) to assess the efficacy of sodium valproate in the

prophylactic treatment of migraine supports the use of this

medication. They employed a double-blind randomized cross-

over design with 29 adult patients. They were given either

400 mg. of valproate b.i.d. or placebo for eight weeks, then

crossed over to the other condition for another eight weeks.

Their results demonstrated that 86.2 percent of the patients

reported effective prevention of migraine or reduction of

frequency, severity, and duration of their migraines. The

drug was reportedly well tolerated (i.e., few adverse

effects). Interestingly, there was no correlation between

the effectiveness of the drug and blood levels. Also,

valproate only appeared to be effective in severe cases of

migraine. Another study (Mattew and Ali, 1991) also showed


a decrease in headache frequency in two-thirds of a group of

patients with intractable chronic daily headaches. They

did, however, report weight gain, tremor, hair loss, and

nausea. Neither study assessed the cognitive effects of

this treatment.


Based on the information provided in this literature

review, several conclusions become apparent. Migraine is a

phenomenon that is not uncommon in children and consequently

is receiving deserved attention from the field of medicine.

The neuropsychological and personality variables are less

well defined in this pediatric sample, allowing for further

research. Another point to be made about migraines is the

many links that have been drawn between migraine and

epilepsy. These include similarities on EEG, the presence

of aura, and familial components in some cases. An

additional intriguing similarity is the use of antiepileptic

drugs in both populations. While the use of these drugs in

the treatment of pediatric migraine has been challenged by

some, other research has shown them to be efficacious.

However, as has been shown in the epilepsy literature,

antiepileptic agents have potential drawbacks.

Phenobarbital, for example, has been shown to produce

sedative effects that may negatively impact attention and

concentration. Other drugs in this class, however, have

proved to be effective without such drastic side effects.

Valproate may have the least negative impact on

neuropsychological functioning, although the research on

this drug is severely lacking. One possible side effect of

valproate to be considered is that children to whom it has

been administered are at risk for carnitine deficiency and

subsequent hepatotoxicity. While the cognitive effects of

such a deficiency are not well described in children, at

least one study has shown that carnitine supplementation

improved cognitive functioning in a geriatric population.

Because most of the research in the area of cognitive

effects of antiepileptics in children has focused on

children with epilepsy, the confound arises that such a

population may have pre-existing differences in brain

functioning, and are therefore not an adequate population to

study in isolation. Migrainous children, then, are

potentially an excellent comparison group. Unfortunately,

the cognitive research on valproate and carnitine

supplementation in children is nonexistent, creating a need

for a well designed study with appropriate controls.

It would be necessary that such a study utilize

sensitive neuropsychological testing so as to describe even

the most subtle effects of both valproate and carnitine.

The questions to be answered include the following:

1. Are there any differences in personality or
neuropsychological functioning between migraineurs and
normal controls?

2. What are the cognitive effects of valproate?


3. What effect does carnitine supplementation have on
neuropsychological measures?




Subjects were recruited from the pediatric clinic at

Shands Hospital as part of a drug study to examine the

treatment efficacy of valproate and carnitine in children

diagnosed with migraine. Accrual of subjects took place

over a 12 month period, beginning in January of 1993 and

ending in December of 1993. The mean age was 12.76 years

with a range of 9.17 to 15.83 years of age. The sample

consisted of 3 females and 11 males. An IQ screening

ensured that all subjects fell within a normal range of

intelligence (subjects with a mean scaled score below 8 or

any scaled score below 6 were excluded from the study). A

total of 14 subjects were recruited into the study and

administered baseline testing. Due to attrition, the sample

size decreased to 13 by the second testing, and to 6 by the

third testing. All subjects in the migraine group were

declared to be neurologically unimpaired by the pediatric

neurologist. EEG data for these subjects was read as


In addition to the migraine group, twenty control

subjects were recruited from area elementary and secondary

school, and a church youth group. Eighteen subjects

remained by the third testing. The age range for this group

was 7 years to 17, with a mean age of 12.08. The sample

consisted of 6 females, 14 males. There was no significant

difference in IQ between the control and migraine groups.

Materials and Apparatus

The following is a descriptive list of

neuropsychological tests used in this study:

Intellectual testing

Four subtests from the Wechsler Intelligence Scale for

Children-Revised (WISC-R) were used as an estimate of

intelligence. From the Verbal subtests, Comprehension and

Similarities both offer high split-half reliability (.77 and

.81, respectively) (Sattler, 1988). Similarities

correlates at .72 with Verbal IQ, while Comprehension

correlates .68. These particular subtests were chosen

because they offer high reliability across the age group

used in this study and are potentially less affected by

educational factors such as socioeconomic status and

educational level. Similarly, Block Design and Object

Assembly also offer good split-half reliability (.85 and

.70, respectively), and contribute information from the

Performance oriented side of the WISC-R. Block Design

correlates with Performance IQ .68, while Object Assembly

correlates at .60.

Together, these four subtests provided a good estimate

of IQ and served as a screening instrument, baseline, and

provided a description of the population during the initial

testing of all subjects. Because IQ was used to ensure a

normal range of intelligence for the subjects, these

subtests were administered only at baseline. The following

tests, however, were administered at each of the three

testing points.

Memory tasks

Verbal Memory

1. Buschke Selective Reminding Test (SRT): This test

of verbal memory is designed to measure verbal learning and

memory during a multiple-trial list-learning task. Each

subject was read a list of 12 words to remember, and then

was required to recall as many of the words as possible. On

subsequent trials, only words the subject failed to recall

on the previous trial were presented again. This format

offers the advantage of assessing which words are encoded

quickly (those that do not require additional prompting)

versus those that are difficult to encode. In addition, it

offers a measure of short term and long term memory, as well

as learning rate. According to Buschke and Fuld (1974),

this test maximized learning by directing the subject's

attention to those words not recalled on the previous trial.


Additionally, recall of an item without its presentation is

interpreted as evidence for encoding into long-term storage

(Levin, Benton, and Grossman, 1982), whereas short term

recall is reflected in items dependent on presentation.

The SRT has a number of different versions of the test,

thus making it ideal for repeat testing. Three versions

were used in this study. Clodfelter et al., (1987) state

that the different forms for children are roughly of

equivalent difficulty. Test-retest reliability is best for

the consistent retrieval aspect of the SRT at .92 (Masur et

al., 1989). Several dependent measures were obtained from

the SRT: immediate span, learning (total of items recalled

from trials 3 through 8), mean number of items recalled per

trial, and delayed recall.

2. Stories from the Wechsler Memory Scale: A modified

version of the Wechsler stories (Logical Memory Form I and

Form II) (Wechsler, 1945) was used to assess immediate and

30-minute delayed memory. There are two stories of

approximately equal length from each alternate form of the

WMS. For purposes of this study, only one of the stories

was administered at any one testing time.

Each form of the WMS contains two different stories.

Both stories from Form I ("Anna Thompson" and "American

Liner") and the first story from Form II ("Dogs of War")

were used, in the order listed here. Although the two forms

of the WMS are not identical, the Logical Memory subtests

are considered to be sufficiently comparable (Spreen and

Strauss, 1991) and are of similar length. Because of

extensive research on reliability and validity as well as

detailed scoring and appropriate norms (Russell Revision),

these stories are a suitable choice for this study. Raw

scores were converted to z-scores by first doubling the raw

value (because norms are based on administration of two

stories), subtracting the normative mean for the subjects

age, and dividing by the standard deviation.

3. Digit Span: In Digit Span, the subject listened to

a series of orally presented digits and then repeated them

in the same order. This measure of short-term memory and

attention has good reliability (.78), but is not predictive

of general intelligence (Sattler, 1988). It is not

significantly affected by practice, and is therefore

appropriate for multiple testing. Normative z-scores were

obtained in the same manner as described above.

Nonverbal Memory

1. Corsi Cubes: Corsi Cubes, like digit span,

requires the subject to repeat a sequence presented by the

examiner. However, instead of verbally presented digits,

the examiner taps cubes, that are one and a half inches in

dimensions, that are randomly ordered on a board. The

subject must then tap the blocks in the same order, or in

reverse order, depending on the examiner's instructions.

Because of the absence of verbal stimuli, it is considered

to be a test of nonverbal memory with spatial components.

It is no more susceptible to practice effects than digit

span, and is therefore appropriate for repeated assessment.

Z-scores were obtained for forward and backward span,

providing two dependent measures.

2. Visual Learning from the Wide Range Assessment of

Memory and Learning (WRAML): The Visual Learning subtest

from the WRAML has a median reliability coefficient of .88,

while test-retest reliability is .81 (Sheslow and Adams,

1990). On this test, the child was presented with visual

designs which are arranged in particular positions on a

board. The child had to remember the spatial location of

the design, while the design is covered by sponge shields,

over a period of four trials. The color and aesthetic

qualities of this task make it ideal for children. Based on

its reliability and nonmeaningful stimuli, repeated testing

should not produce significant practice effects. A scaled

score based on the total number of correct responses was

obtained using the age-appropriate norms provided by the

WRAML manual.

Motor Tasks

1. Finger tapping: This task required subjects to

place their hand on a board with their index finger

positioned over a metal key. When depressed, this key

advances the numbers on a counter so that a tapping rate per

ten second interval may be obtained. Each hand was tested


on four trials each, and the rate was averaged across trials

to yield a number for each hand, which was then translated

to z-scores. Brown, Rourke, and Cicchetti (1989) conducted

research on the reliability of different neuropsychological

tests for children. Although they claimed to have found the

reliability for finger tapping to be excellent (.60) for the

dominant hand, and good (.58) for the nondominant hand,

these numbers are modestly reliable. The use of a control

group was intended to control for extraneous factors.

2. Beery Test of Visual Motor Integration (VMI): The

VMI is designed to measure perceptual-motor abilities by

requiring the subject to copy up to 24 geometric forms.

Reliabilities for the VMI range from .81 for test-retest, to

.93 for interrater reliability. Sattler (1988) comments

that the validity is satisfactory, and though the

standardization sample included 3,090 children, it is not

well defined. This test is not likely to be significantly

affected by environmental factors such as medication, and

was chosen to offer a useful comparison to other, more state

oriented measures. It was administered to subjects only at

the first and last testing. An age-based standard score was


Attentional Tasks

1. Computerized Continuous Performance Task (CPT):

The CPT can be of two types: vigilance or reaction time.

One common reaction time format is the single-choice


reaction time (RT), which requires the subject to press one

of two keys immediately after cue presentation. For

example, if the letter X appears on the screen, the subject

presses a key with their left hand (such as the letter Q on

a standard key board). If the letter Y appears, they must

press a key with their right hand (such as the letter P).

The stimuli are presented at random points on the screen.

This format was used in this study to assess reaction time

changes with the administration of valproate.

The value of such a task is that some degree of speeded

decision making must take place in the context of sustained

attention in order to perform correctly. The added element

of having to choose between two responses modestly increases

the level of difficulty, therefore creating a more sensitive

test measure. In a study by Fennell, Fennell, Carter, Mings,

Klausner, and Hurst (1990), this type of task was used

effectively to show differences between a renal pediatric

population and controls. Because it is computer

administered, repeated presentations can be identical or

varied slightly (such as changing the stimulus letters) to

control for learning effects. The motor output is minimal,

leaving the effects of speeded decision making and attention


The CPT is programmed on an IBM computer (Fennell, et

al., 1990). For purposes of this study, the program was run

on either a Leading Edge portable computer to accommodate

change in testing site, or a standard IBM computer. The

task was a simple-choice reaction time in which the subject

had to respond to the letter X by pressing the Q key on a

keyboard, and to the letter Y by pressing the P key on the

keyboard. The letters were presented singly (i.e., one X or

one Y) at random spatial locations on the screen, at random

intervals. The intertrial intervals alternated between 2,

5, and 8 seconds. Exposure time alternated between .3, .6,

and .9 seconds. Each of the 108 trials was repeated twice

for a total of 216 presentations. Correct responses were

those where both of each repeated trials were responded to

correctly. Mean reaction time across all trial pairs and

conditions served as the primary dependent measure for this

task. No normative data is available for this task,

necessitating the use of covariate statistics in analysis.

2. Paced Auditory Serial Addition Test (PASAT): The

PASAT is a serial addition task typically presented via

audio tape. This method of administration controls for rate

of presentation and difficulty level, which can be 2.4, 2.0,

1.6, or 1.2 seconds per digit, with a total of 61 digits per

trial. The subject is required to listen to a series of

digits and, upon hearing the second, add it to the first,

then adding the second to the third number, and so forth.

For example, upon hearing "3, 7", the subject should answer

"10". If the next digit is "5", the subject should arrive

at "12", by adding 7 and 5.

This test reportedly is a sensitive measure of

information processing and sustained attention (Gronwall and

Wrightson, 1981). Gronwall and Sampson (1974) describe the

PASAT as a test that assesses some central information

processing capacity analogous to that seen on reaction time

and divided attention tasks. Because of its difficulty and

attentional demands, it is an ideal test to measure subtle

changes in the ability to process information at a given


One drawback to the PASAT is the evidence of

significant practice effects (Gronwall, 1977). However,

after the second presentation, these practice effects are

not a factor as performance does not continue to improve at

such a rate. For the second presentation, though, a control

group is requisite to compare practice effects and tease out

their impact versus the impact of medication. Spreen and

Strauss (1991) report excellent split-half reliability (.96)

for this measure.

Subjects were administered the 2.4 and 2.0 second

interval trials. A total of 49 correct responses for each

of the two conditions was possible. Normative data allowed

for z-score transformation.

Parent Report

Child Behavior Checklist (CBCL): A commonly used

parent report child assessment tool is the Child Behavior

Checklist (Achenbach & Edelbrock, 1986). The CBCL can be

self-administered to the child's parent or guardian. The

CBCL provides separate norms for 3 age groups: 4 to 5, 6 to

11, and 12 to 16 by sex. There are subscales within each

age group, as well as the two broad-band factors of

Internalizing and Externalizing. Children who rate as

externalizers may be seen as more overactive and as

possessing more conduct problems than those who rate as

internalizers. Factor loadings from the narrow-band scales

justified the labels "Interalizing" and "Externalizing."

The CBCL is very well standardized and has adequate

reliability and validity. It is reported by DuPaul et al.

(1991) to be the best standardized rating scale in use.

Due to the number of variables and relatively small

sample size, the data from the CBCL was summarized into an

Internalizing score and an Externalizing score (both scaled

as t-scores). A third variable was a simple bipolar

assignment which was made for each subject according to

whether they did or did not have any scale scores that were

in a clinically significant range (i.e., 1 70).


A screening battery was administered to all subjects

prior to their introduction to medication. This battery

included all test instruments listed above, including the IQ

screen and parent report measures. In the second and third

testing, however, IQ was omitted, and the CBCL and VMI were

omitted from the second testing. Following baseline


testing, all migraine subjects were then given valproate for

the period of four weeks, after which they were retested to

assess any changes in neuropsychological functioning as a

result of the valproate. Valproate dosage was established

according to weight of the subject, and was titrated

according to blood levels (goal therapeutic range was 50 to

100 ml). Following the second assessment, subjects were

randomly assigned to one of three groups: 1) Group 1

continued on valproate and also receive carnitine; 2) Group

2 continued on valproate and received a carnitine placebo;

and 3) Group 3 received placebos for both valproate and

carnitine, and was therefore drug free. Subjects remained

in their respective groups for a period of 8 weeks, after

which they were tested for a third and final time (see

Figure 1).

Test order was constant across testing. For the

initial testing, the IQ screening will be administered,

followed by a 10 minute break. During this break, the

subtests were scored to determine if the subject met the

minimum criterion for admission to the study. This break

also served to minimize any fatigue effects of the IQ screen

before the actual test battery was administered. The

remaining tests were given in the following order: WMS

story (immediate), CPT, PASAT, WMS story (delay), Finger

Tapping, SRT learning trials, Visual Learning (WRAML), Digit

Span, Corsi Cubes, SRT delay, Visual Learning delay, and VMI

(the VMI only on first and third testings. The CBCL was

completed concurrently during the first and third testing.

During the initial four week period when all subjects

were taking valproate, only the subjects were blind.

However, when the subjects were randomized into one of three

groups, both subjects and experimenters were blind to



(1) There will be no significant differences on

neuropsychological tests between the migraine children and

the control group at baseline.

(2) Parent report (CBCL) will reveal subtle

differences on scales such as somatic complaints and

anxiety. This difference is not likely to be significant

given the effect and sample size.

(3) Following one month of valproate therapy, the

experimental group will perform worse than the nonmigraine

control group on neuropsychological tests that are most

sensitive. Attention oriented tasks (CPT reaction time and

the PASAT) and memory (verbal and nonverbal) tasks will be

the most likely to reveal differences between the groups.

(4) After dividing the experimental group into three

subgroups, the group who is receiving carnitine supplement

will perform better on sensitive neuropsychological tests

than those who are not receiving the supplement, although

this difference may not be statistically significant.

However, there will be a significant difference in

performance such that the third group (those on valproate

placebo and carnitine placebo) will show significantly

better performance than the other two groups. This implies

that valproate will have the most negative impact on

cognitive functioning. Finally, none of the experimental

groups will perform better than the nonmigraine control


(5) After receiving treatment for their migraines,

there will be a decrease in parent report of behavioral

problems, such as anxiety or somatic complaints.



12.8 (2.2)

Mean Age


12.1 (3.3)

Age Range

Female/Male Ratio

IQ subtest means

Block Design
Object Assembly

Number of Subjects
at baseline





11.4 (2.2)
10.2 (2.2)
9.2 (2.6)
10.4 (4.0)

12.3 (3.5)
13.1 (2.9)
11.1 (2.3)
11.5 (3.7)

NOTE: Standard deviations are provided in parentheses.



There was no significant overall group effect for IQ.

All subjects were considered to be within a normal range of


Descriptive statistics were conducted for both groups

at baseline. Table 3-1 provides the means and standard

deviations for both groups separately and as a whole for all

variables used at baseline. Table 3-2 provides the same

statistics for scores at the second testing.

The first hypothesis for this study was that children

with migraine would not differ from the control group at

baseline measurement. To assess the validity of this

hypothesis, a multivariate analysis of variance (MANOVA) was

conducted on the domain scores for each subject at baseline

to determine if there was any significant difference between

migraineurs in an unmedicated condition and controls.

Specifically, items related to verbal memory constituted one

domain and was analyzed together, as were nonverbal memory

items. Domains analyzed were as follows: Verbal Memory

(Logical Memory and the Buschke SRT), Nonverbal Memory

(Visual Learning from the WRAML), Memory Span (Digit Span

and Corsi Blocks), Attention (PASAT and the CPT reaction


time), Visual-Motor (VMI), and Behavior (CBCL). By grouping

the variables into MANOVAs as described, the number of

statistical analyses was reduced in an attempt to control

for a Type I error. For the Attention domain, a MANCOVA

(multivariate analysis of covariance) with age as the

covariate was required because of a lack of normative data

for the CPT. The study by Fennell, et al. (1990)

demonstrated that there were age effects for reaction time,

such that reaction time decreased as age increased. The use

of a MANCOVA was intended to take this age effect into

account for these analyses.

At initial testing, no significant differences were

seen between the migraine and control groups on any of the

neuropsychological tests. A t-test to assess any

differences on the VMI approached significance (p = .052),

with the migraineurs reflecting a lower mean score for the

VMI than the control group.

The second hypothesis was that the parents of

migrainous children would endorse more problem behaviors,

particularly those considered to be a reflection of somatic

complaints, than parents of the control children. Scores

from the CBCL were summarized into t-scores for

externalizing and internalizing factors. The mean

internalizing score for the migraine group was 61.85, and

the mean for the externalizing score was 55.08. Because too

few parents returned the CBCL for the control group, the

migraine group scores were compared to the published

normative data using a Welch's T-test. For the normative

data, mean scores are 50 with a standard deviation of 10.

Using this method, a significant difference resulted for the

internalizing factor (p < .05) but not for the externalizing

factor. The mean number of clinically elevated scales for

the migraine group at baseline testing was 2, with a range

of 0 to 9. The most commonly elevated scale was Somatic


The third hypothesis tested was that subjects taking

valproate after one month would perform significantly worse

on measures of memory and attention than the non-medicated

control group. A repeated measures MANOVA was conducted for

each domain for the first and second testing to assess the

effect of introducing valproate to the migraine group.

Results from this set of analyses resulted in no significant

group effect for any of the neuropsychological domains

measured. The repeated measures MANOVA for Span was the

only domain analysis that approached significance (F = 2.46,

p = .069).

After examining the blood levels of valproate for the

migraine group, it was apparent that many of the subjects

were not in a therapeutic range (i.e., 50 or higher).

Therefore, another set of repeated measures MANOVAs was

performed with different groupings. Because there were no

significant differences observed at baseline between the

migraineurs and controls, those migraineurs who were not

considered to have a therapeutic blood level of valproate

were re-assigned as controls for this analysis. Therefore,

the groupings resulted in those migraineurs with a

therapeutic blood level of valproate versus all other

subjects. The results from this analysis yielded a

significant group by time interaction on only the Attention

domain (F = 5.06, p = .036). This was actually a repeated

measures MANCOVA, with age as the covariate due to a lack of

age norms for the CPT reaction time data. The results

indicate that group membership and time of testing (i.e.,

after one month of valproate therapy) were important

variables on the Attention domain.

Difference scores were obtained between the variables

on testing one and two, and a correlation analysis was

conducted on these difference scores and the blood levels

obtained at the second testing. This analysis therefore

includes only migraineurs. Although none of the

correlations obtained were statistically significant,

several did approach significance. Both the fast and slow

trials of the PASAT were close to significance (p = .083 and

p = .058, respectively), but in opposite directions. The

fast trial of the PASAT revealed a positive correlation

while the slow trial revealed a negative correlation. The

immediate recall of Logical Memory also neared significance

(p = .064, positive correlation), as did Visual Learning

from the WRAML (p = .062, negative correlation). These may

have reached significance with a larger sample (total sample

size for these correlational analyses ranged from 11 to 13).

The fourth hypothesis involved data from the third

testing, after migraine subjects were randomly assigned to

one of three groups (valproate plus carnitine, valproate

plus placebo, and placebo plus placebo). The results from

the third testing were unable to undergo statistical

analysis because of a high attrition rate resulting in an

extremely small sample with group sizes of one to three

subjects. Correlations between blood level and the third

testing did not result in any significant relationships,

perhaps due to the even smaller sample size.

The fourth hypothesis also predicted that carnitine

would improve cognitive functioning in those subjects

receiving valproate, but this difference would not be

statistically significant. This could not be analyzed due

to the severe attrition. Additionally, there was no

equivalent check for compliance as there was for the

valproate as measured by blood level. Therefore, even those

subjects assigned to take carnitine were not clearly under

the influence of the supplement.

The fifth hypothesis was that migraineurs would show

improvement by the third testing with the administration of

valproate on parent report of problem behaviors. The CBCL

data was analyzed for the seven subjects remaining at the

third testing. Neither the internalizing or externalizing

scores differed significantly from the normative data.

Although this is consistent with the hypothesis, the

internalizing score remained higher than the externalizing

score (64.29 compared to 59.14), indicating that problems

behaviors in this domain were still perceived by parents to

be more problematic than the externalizing behaviors.

Failure to reach significance is probably due to sample

size, given that the means are slightly higher than at

baseline. The increase in means from baseline to third

testing is likely a result of differential attrition, but

this was also not significant. Also, not all subjects were

in a therapeutic range of valproate because of group

membership and due to compliance issues.

Post-hoc Analyses

A significant practice or test order effect (F = 7.85,

p = .01) was observed for all subjects for the Visual

Learning test that comprised the Nonverbal memory domain, as

well as Logical Memory (p = .000).

A Chi-Square was performed to determine if group

membership among the migraineurs (i.e., placebo versus drug

conditions) predicted attrition. From the valproate plus

carnitine group, 75 percent dropped out, while 57 percent

attritted from the valproate plus placebo group, and 25

percent attritted from the placebo plus placebo group.

Although these numbers appear to be significant, Chi-Square

analysis did not support this (p = .35). This lack of

statistical finding may also be due to a relatively small

sample size.

A test of binomial proportion was conducted on two

groups (the two groups administered valproate were collapsed

into one group and compared to the placebo plus placebo

group). The corresponding rates of attrition for the groups

were 70 percent for the valproate group, and 25 percent for

the placebo group. This approached significance (p = .062).

To determine if adolescents were more likely to be

noncompliant than children, the migraineurs were divided

into two groups: those 13 and older, and those 12 and

younger. A t-test based on these two groups and comparing

the mean blood levels after one month of valproate therapy

resulted in a significant finding (p = .002), such that

adolescents had a lower mean blood level. Among the seven

adolescents in this analysis, five had blood levels that

were not in a therapeutic range (i.e., less than 50). All

of the children (ages 12 or younger) had blood levels over


Finally, to determine if compliance was related to

socioeconomic status, the subjects were divided into those


that were labelled medically needy and receiving aid from

Children's Medical Services, and those who had private

insurance coverage. This did not result in a significant

difference in blood level between the two groups.

Table 3-1
Means and Standard Deviations for Migraine and Control Group
at Baseline



(The following group means are based on scaled scores)


10.2 (2.66)

9.5 (2.20)

10.7 (2.90)

(The following group means are based on z-scores)

Logical Memory
Logical Memory
Digits Forward
Digits Backward
Corsi Forward
Corsi Backward
Motor Tapping

CPT Reaction
VMI Standard
CBCL Intern.
CBCL Extern.

.84 (1.66)


( .61)
( .53)
( .54)
( .62)
( .97)

40.19 (9.20)

99.10 (9.52)

.74 (1.98)


( .39)
( .42)
( .57)
( .66)
( .98)

.93 (1.35)


41.64 (10.0)

94.38 (6.80)

61.85 (9.95)*

55.08 (11.8)*

( .68)
( .61)
( .54)
( .60)
( .98)
( .89)
( .87)

39.2 (8.81)

102.0 (10.0)

* The migraine group was compared to the published normative
data for these measures. Therefore, no means are listed for
the control group.


Table 3-2
Means and Standard Deviations for Migraine and Control Group
at Second Testing



(The following group means are based on scaled scores)


11.55 (3.25)

10.92 (3.45)

11.95 (3.14)

(The following group means are based on z-scores)

Logical Memory
Logical Memory
Digits Forward
Digits Backward
Corsi Forward
Corsi Backward
Motor Tapping

CPT Reaction
Time 4

-.63 (1.40)


( .65)
( .72)
( .65)
( .68)
( .96)

17.94 (14.9)

-.83 (1.72)


-.50 (1.17)

( .38)
( .93)
( .58)
( .47)
( .81)


50.44 (21.5)

( .68)
( .56)
( .68)
( .79)
( .97)

47.0 (12.7)

(The following scores were obtained from the third testing)

VMI Standard
CBCL Intern.
CBCL Extern.

106.2 (10.7) 102.20 (13.8)

107.4 (9.77)

64.29 (12.5)*

59.14 (14.24)*

* The migraine group was compared to the published normative
data for these measures. Therefore, no means are listed for
the control group.




The literature is represented by heterogeneous results in

the studies of both migraine and valproate. The underlying

mechanisms of both continue to be disputed. In addition to

the discrepant findings on the neuropsychological effects of

migraine in adults, the introduction of childhood and

adolescent migraine further clouds the picture. Several

authors have claimed to have found neuropsychological

differences between adult migraineurs and controls (e.g.,

Kohler & Fennell, 1992; Crisp, Levett, Davies, & Rose, 1989;

and Hooker & Raskin, 1986), while others assert that no

differences exist (e.g., Leijdekkers, et al., 1990; Burker,

Hannay, & Halsey, 1989). Their discrepant findings may be the

result of a number of different factors, including

inappropriate statistics conducted on relatively small sample

sizes, leading to spurious results. Other factors may include

the use of cognitive tests that paint the neuropsychological

picture in broad strokes rather than using more sensitive

measures that target subtle cognitive changes. Additionally,

the ages of the subjects and migraine history may play a key

role in the findings. Regardless of the adult studies on

migraine and cognitive functioning, there remains to be


addressed the question of cognitive differences in children or

adolescents with migraine. They may comprise a different

subset of migraineurs due to the necessarily different history

of the migraine process.

The debate over the neuropsychological impact of

antiepileptic medications has also yet to be resolved. Older,

supposedly more detrimental medications such as phenobarbital

have been examined in several studies and their cognitive

impact has been described. However, the effects of valproate

remains controversial regarding whether patients receiving the

medication suffer a compromise in certain aspects of

neuropsychological functioning. One problem in assessing this

medication is the population used to test its effects.

Historically, adults or children with epilepsy have been a

favored sample on which to demonstrate the effects of

valproate, among other similar medications. The results have

been mixed with this method (e.g., Aman, Werry, Paxton, &

Turbott, 1987; Bittencourt, Mader, Bigarella, & Doro, 1992;

Blennow, Heijbel, Sandstedt, & Tonnby, 1990) and the methods

employed have included both the administration of valproate in

polytherapy, and the withdrawal of medication in presumably

stable epilepsy patients. The drawbacks to this population

are obvious: epilepsy patients may have poorer

neuropsychological performance due to the epilepsy rather than

to the medication. It is unethical to place a patient with

epilepsy in a placebo control group, therefore making the

distinction between the epilepsy and drug effect uncertain.

Withdrawal studies also due not control longer lasting effects

of medications well, although blood levels may control for

some of this effect.

Additional studies of the effects of valproate have

included normal adult volunteers to assess the effects of the

medication (e.g., Harding, Alford, & Powell, 1985; Thompson &

Trimble, 1981). This method is also not without its

drawbacks, however. Most studies of this type only administer

the medication for short periods of time, often without a

blood level check to ensure therapeutic levels. The few

studies of this type are also plagued with methodological

problems such as poor assessment of neuropsychological

abilities and inappropriate statistical procedures. Although

this method may prove beneficial under the appropriate

conditions, to date these results have not yielded

substantial, useful information. Additionally, it is a method

that would preclude the use of normal children in most cases,

therefore omitting important data regarding this potentially

different group.

Several studies have documented the efficacy of valproate

as a prophylactic medication for migraine (e.g., Sorensen,

1988; Hering & Kuritzky, 1992). This provides another

population useful for studying the cognitive effects of the

medication, particularly in children. Although some studies

have determined that migrainous individuals may have some

different neuropsychological deficits, this has yet to be

demonstrated in children. If in fact, migrainous children are

not significantly different on cognitive measures that other

children without medical compromise, then they would provide

an excellent opportunity to assess more pure effects of


Interpretation of Findings

Several important findings result from the current study.

First, no significant differences were found on baseline

testing measures between the migraine group and the control

group. This study is one of the first to examine the

potential for neuropsychological differences in an adolescent

migraine group. Results suggest that there are no

neuropsychological differences between children and

adolescents with migraine and a similar group of controls.

One exception to this is the difference between groups on the

VMI (p = .052). Although this finding suggests differences in

visual motor integration, this finding was not apparent at the

third testing, perhaps due to the decreased sample size.

Overall, the finding of no difference at baseline unmedicatedd

state) is in contrast to a few of the adult studies that have

found differences on cognitive measures. One potential reason

for this is differences in methodology and statistical

procedures. Another interpretation, however, is that a

fundamental difference exists between adults and children with

migraine. Some have suggested that migraine, like epilepsy,

compromises the brain through repeated vascular attacks

(Farkas, Kohlheb, Benninger, & Matthis, 1992; Leviton,

Malvea, & Graham, 1974; Eviatar, 1981) that may have long-

term, permanent effects. Therefore, patients with a longer

history of migraine would be more likely to exhibit cognitive

impairments. Children and adolescents with migraine, however,

may not have the extended history of migraine necessary to

demonstrate such impairments. Additionally, the fact that

children are less likely to be diagnosed with classic migraine

than adults may also play a role in the cognitive differences

between these populations.

Despite a lack of differences found on neuropsychological

measures, migraineurs did demonstrate higher internalizing

problem scores on the CBCL than is expected from the general

population. Externalizing behaviors were not reported as

being significantly different. These results are not

surprising given the frequency of scale elevations for Somatic

Complaints, which contains report of headache. However,

several subjects had more than one scale elevation, which may

reflect personality differences in adolescents with migraine

versus their headache-free peers. This is consistent with the

research on adult migraine (e.g., Leijdekkers, et al., 1990;

Burker, Hannay, & Halsey, 1989) that has supported differences

in personality even without evidence of neuropsychological

differences. Research examining personality characteristics

of adolescent migraineurs (e.g., Andrasik, et al., 1988;

Cooper, et al., 1987) has found higher reports of depression

(which is considered to be an internalizing behavior on the

CBCL), among other behaviors, and supports the findings from

this study.

Analyses conducted on the first and second testing

indicated that valproate has little effect on most

neuropsychological domains with the exception of attention.

This finding was only borne out when the subjects were re-

grouped according to therapeutic blood level, which lends

further support to the finding that migraine group membership

alone does not reveal differences. It may be that the

measures of attention used in this study are more sensitive to

subtle alterations in functioning caused by valproate than the

other measures. Attention itself may be more sensitive to

external factors than memory, although greater degrees of

impairment in attention are likely to have a negative effect

on memory. The findings by Hwang and Van Woert (1979) that

valproate has a greater effect on striatal and hippocampal

areas of the brain supports the differences found in attention

in this study because of striatal connections with frontal

lobe structures, but does not explain the apparent lack of

results regarding memory. Again, varying degrees of test

sensitivity may be the cause.

Unfortunately, the original design intended could not be

analyzed due to the degree of attrition. The third testing

was originally intended to measure the effects of valproate


with placebo controls in a double-blind manner, as well as the

concurrent administration of carnitine. With only three

subjects remaining that were taking the carnitine supplement,

statistical analysis was impossible. Additionally, these

subjects demonstrated a level of noncompliance with the

valproate that could be measured with blood levels, but this

method was not possible with carnitine. Therefore, of the

three subjects assigned to carnitine, all or none of them

could have actually been taking the supplement. Analysis,

even qualitative, would be rendered meaningless. Although not

significant, a greater number of subjects in the valproate

plus carnitine group dropped from the study, with the

valproate plus placebo group comprising the next largest group

of drop-outs, followed by the placebo plus placebo group. The

negative side effects of the medication may have been the

cause of this differential attrition, as well as the aversive

taste of the carnitine (one subject confided that he was not

taking the carnitine for this reason).

Adolescent Noncompliance

Because compliance was such a significant factor in this

study, this topic deserves further discussion. The literature

on compliance in adolescents and children has increased

significantly over the past two decades. In an extensive

review of the compliance literature, Dunbar-Jacob, Dunning,

and Dwyer (1993) noted that 91 studies were conducted in this

area between 1970 to 1989, with research ranging from simple


assessment of compliance to intervention or treatment designed

to increase compliance. However, of the 91 studies conducted

during this time span, none included specific assessment of

the compliance of migrainous adolescents. Also, a search for

research employing such a population between 1989 and 1993 was

unsuccessful. Despite the lack of research on a migrainous

population, the results of numerous other studies on

compliance in adolescents are helpful in attempting to

understand better the underlying causes of the low compliance

in this study.

The literature on childhood compliance in general reveals

dramatically poor adherence to medication regimens, even with

potentially life-threatening illnesses such as cancer. This

low level of compliance has been shown to be worse for

adolescents compared to younger children (e.g., Tebbi, 1993;

Johnson, 1993). In a study specifically designed to assess

the level of compliance in adolescent patients with acute

lymphoblastic leukemia or Hodgkin disease, 52 percent of the

patients were noncompliant with their prednisone therapy,

while nonadherence was demonstrated in 48 percent of those

prescribed penicillin for post-splenectomy prophylaxis. These

findings are not unusual, which is demonstrated by the

findings of Ettenger et al. (1991). They found a medication

noncompliance rate of 50 percent in 70 pediatric renal

transplant patients. The rate was slightly higher for

adolescents (64 percent). Unfortunately, the result of this

noncompliance was a 13 percent loss of transplant graft.

Given that the literature has supported the belief that

adolescents generally demonstrate poor compliance with medical

regimens, the focus then should be on what factors contribute

to poor compliance and how compliance might be improved among

this group. Brooks-Gunn (1993) detailed several

biopsychosocial risk factors influencing adolescent health

behavior. One of these factors is the cultural context in

which the adolescent is embedded. Different cultural views on

medical adherence and the locus of responsibility (i.e., who

is responsible for the administration of medications) can

impact the level of compliance achieved in a given population.

In addition to cultural factors, individual factors also play

a significant role. These include biological (e.g., hormonal

changes and levels of depression), emotional, and social-

cognitive (e.g., perceived costs and benefits, risks, and

future consequences) factors. These issues are particularly

relevant in the adolescent population who are often seen as

being naive to consequences of risky behavior. Environmental

factors may also play a role in compliance, particularly via

the influence from peers and parents. As children enter

adolescence, conformity becomes increasingly important.

Perceived differences between the patient and peers based on

medication usage and side-effects may contribute to

noncompliance. An important variable is the parental

influence, which may become less as the adolescent becomes

increasingly independent. In fact, it has been stated that

parents themselves become less compliant with medical regimens

as their children enter adolescence (S. B. Johnson, personal

communication). A final environmental factor is that of the

physician/patient communication. The patient may feel that

they are being compliant, but due to poor communication, they

may actually have gaps in their knowledge about the medication

or treatment plan that leads to inadvertent noncompliance

(Johnson, 1993). This is most likely to be problematic for

younger children, but can also be seen in adolescents.

In a study of the psychosocial determinants of compliance

in adolescents with iron deficiency, Cromer, Steinberg,

Thornton, and Shannon (1989) reported a 67 percent compliance

rate. Factors they found to be statistically significant

predictors of compliance included side effects of the

medication and frequency of family reminders to take the

medication. This suggests that, even with adolescents,

instruction to the parents to become more involved in their

children's medical adherence may be beneficial. Also, the

report that negative side-effects tended to decrease

compliance is consistent with the descriptive results from

this study, in which several patients complained of negative


The concordance of parent and adolescent views on

medication has been explored in at least two studies. Tebbi,


Zevon, Richards, and Cummings (1989) measured cancer patients

and their parents on attributions related to their degree of

perceived responsibility. These authors failed to demonstrate

a relationship between parent-child concordant attributions

and medical compliance. However, in an earlier study, Tebbi

et al. (1988) did find that compliance was greater when

parents and patients agreed on who was responsible for

administration of the medication and their understanding of

instructions and effectiveness of the medication. This would

theoretically alter the diffusion of responsibility (which

would likely reduce adherence) by placing the responsibility

on one individual. However, even in such a model, parental

support would likely benefit the adolescent.

Noncompliance has also been associated with a restriction

of the adolescent's independence, low self-esteem, and poor

family relations (Friedman, 1986). One method of increasing

compliance, therefore, would appear to be increasing self-

esteem and perceived independence, as well as clarifying to

both the parent and adolescent who is primarily responsible

for the administration of the medication. This is consistent

with the research of Tebbi et al. (1988). Weinberger (1987)

noted that compliance can also be enhanced by including the

patient in decision making in order to foster an internal

locus of control. These suggestions appear to be relatively

simple methods of improving compliance among adolescents.

Assigning a health paraprofessional or research coordinator to

assess compliance throughout a study, improve communication,

and offer reinforcement for adherence, may also be beneficial.

The level of noncompliance in this study is probably

multi-factorial. The fact that adolescents were significantly

less compliant than the children is consistent with the

literature. Also, the majority of compliance research has

focused on children and adolescents with potentially fatal

illnesses such as cancer. The fact that the subjects in this

study were being treated for a relatively benign, although

aversive, problem (i.e., migraine) may have led to even lower

compliance rates. Typically, headache pain is treated by the

sufferer on an as-needed basis with over the counter

medications. Therefore, the idea of long term daily use of a

medication that has potential side effects may have also

contributed to low adherence. Finally, the chaotic nature of

many of the families in the study may have created

difficulties for the patients in their ability to maintain a

consistent schedule of administration. The incentive to

change their migraine situation compared to the negative

aspects of consistent medication may have not been great

enough to lead to good adherence.

The literature on adolescent and childhood compliance is

helpful in taking steps to increase compliance for treatment

and research purposes. In retrospect, several steps could

have been taken to reduce the level of attrition and improve

the value of the study. Closer follow-up with each of the

subjects and their parents, both in clinic and via telephone,

could have served to help the subjects feel more engaged in

the study, and to address any concerns they may have had at

any point. Through frequent contact and better education

about the medication and its side effects, the subjects and

parents may have been less likely to drop out and instead

communicated any problems they were having. This follow-up

was attempted at a lesser level, with reminders for

appointments being made by telephone, but failed to address

the broader issue of communication. Diffusion of

responsibility was also problematic in this study. Those that

created the study were not always the ones in direct contact

with the subjects, and therefore not responsible for follow-

up. Others that had more contact, did not feel responsible

for conducting the follow-up either, resulting in a gap of

service. Assigning one individual specifically to conduct the

follow-up would have lessened the diffusion of responsibility,

and perhaps reduced attrition. Part of the follow-up and

original education could have been designed to engage the

parent in the study, particularly for adolescent subjects.

The issue of diffusion of responsibility most likely operated

at this level as well, with parents and adolescents feeling as

if someone else was responsible for their adherence. By

engaging the parents early in the study, this issue could have

been clarified with the families, with specific

responsibilities assigned to both adolescent and parent.


A methodological change could also have helped circumvent

the attrition issue. Most of the attrition occurred after the

randomization of subjects into three groups, after the month

of valproate treatment. A better design may have been to

randomize the subjects immediately after baseline testing,

thereby obtaining data from more subjects after only one

month, rather than three months when many had dropped-out.

This would provide a placebo control group as well as

carnitine data that was not available in this study. A cross-

over could then occur after one month to obtain a cleaner

picture of valproate's effects. This methodological change

would not, however, remove the compliance problem as it

applies to drug level. This compliance may have been improved

by the methods described above, as well as employing more

frequent blood level checks and appropriate follow-up.

A final method of minimizing attrition is by making the

entire study less aversive. Reducing the amount of time the

subjects had to spend per visit, either by waiting for their

appointment or the time spent in the neuropsychological

testing, may have made subjects less likely to drop-out of the

study. Providing them with an incentive (e.g., monetary) may

also have helped reduce the aversive nature of the study and

decreased attrition.

Limitations of the Study

As has been alluded to above, the level of attrition in

this study created statistical problems that precluded testing

certain hypotheses. Specifically, those measures from the

third testing could not be adequately analyzed due to small

sample sizes. A simple correlation between blood level and

performance among the migraineurs failed to yield significant

findings. Again, the small sample size is most likely

contributory. This prevented an adequate use of a placebo

control group, although a normal control group helped to

distinguish practice effects. It also prevented assessing the

effects of the carnitine supplement altogether. However, with

the modest differences found between a therapeutic level of

valproate group and all other subjects, carnitine would not

likely have been able to have had a significant effect on

cognitive functioning.

The small sample size at baseline (i.e., 34 subjects) may

be criticized as being too small to reveal potentially

important results at baseline, thereby obscuring actual

differences. Although a power analysis may be suggested as

one method of determining whether the lack of findings was due

to a sample size too small for proper analysis, this is not

without controversy. There were statistically significant

findings using repeated measures analysis (i.e., an

interaction of testing time and group) for the Attention

domain. This suggests that the level of power was therefore

adequate, at least for this domain, to detect differences at