Pathology of the dominant hemisphere

The cognitive manifestations of epilepsy primarily reflect the region of pathology from which seizures arise, and a number of key clinical variables modify the cognitive presentation.

An organizing principle of cognitive impairment in seizure disorders is the distinction between pathology of the language-dominant (typically left) and language-nondominant (typically right) hemispheres. As such, determining whether such impairment exists is a key goal of presurgical neuropsychological assessment.

Hemispheric dominance and handedness. Handedness is strongly correlated with hemispheric language dominance, and as such provides an estimate of the language-critical hemisphere. In both left- and right-handed patients, the left hemisphere is most commonly language dominant; in right handers, incidence of left hemisphere dominance is usually estimated at around 95% whereas this drops to around 70% in left-handed patients. For instance, in a pair of particularly large studies of healthy right-handed individuals, 92.5% had left hemisphere language dominance whereas 7.5% had right hemisphere dominance (41). In a more targeted study of atypical dominance, only 4% of strongly right-handed patients had right hemisphere language dominance; conversely, 27% of strongly lateralized left handers and 15% of ambidextrous individuals had right hemisphere language dominance (42).

However, another large study on healthy individuals (n = 297, 159 left handers) reported the relationship between hemispheric dominance for language and for hand to be scarcely above the chance level (53). Using Gaussian mixture modeling in right and left handers, the authors found that the only exception is a rare group of individuals (below 1% of the general population) showing strong right hemispheric dominance for language and hand preference. Concurrently, Johnstone and colleagues who examined 58 left handers and 33 right handers reported that the left handers were less left-lateralized for language than the right-handed participants (40). Importantly, the left handers were considered to have typical fMRI language dominance.

In epilepsy, the rates of atypical language dominance are higher, likely due to early reorganization caused by the epileptic focus, and the relationship to handedness is far less determined (30). Given that interpretation of cognitive data rests on a probabilistic assumption that the left hemisphere supports language, when definitive statements are to be made (eg, presurgical planning) it is essential to confirm language dominance. This can be done using measures such as clinical fMRI of language (13), magnetoencephalography (66), intracarotid sodium amytal (Wada) testing (43), and direct cortical stimulation mapping (06).

Impairment of language in epilepsy. Individuals with epilepsy are more likely to have atypical organization of language function than healthy controls (26). About one fourth of individuals with epilepsy have been reported to have their language organized atypically in the brain (12). The most common predictors of atypical language distribution in the brain are left-handedness (39), location of lesion (60), early seizure onset (86), pathology (eg, hippocampal sclerosis) (68), high seizure frequency (86), and epileptic activity (07).

The nature of language impairment in epilepsy reflects the precise region of the language system from which seizures arise. Our understanding of the brain regions involved in language--and the deficits accompanying their impairment--has rapidly evolved with the advent of noninvasive imaging (eg, PET, fMRI, MEG). Historic models of language function centered on the contributions of Paul Broca and Carl Wernicke, describing a fundamental relationship between “expressive” language and the posterior inferior frontal gyrus (Broca’s area), and between “receptive” language and the posterior superior temporal gyrus (Wernicke’s area) (72). Overwhelming evidence has supported modification of this model to reflect the contributions made by a range of other regions (17; 52). Rapidly growing evidence strongly suggests that complex cognitive functions (such as language) engage networks that dynamically interact with one another, rather than single cortical areas (51).

Epilepsy patients seem to have altered functional connectivity for language. They have a disrupted recruitment of the language network in the temporal lobe that is ipsilateral to the focus of seizures. The magnitude of altered connectivity has been linked to earlier seizure onset as well as longer seizure duration (85; 86). Altered connectivity for language has even been reported in children with new onset benign childhood epilepsy using resting state fMRI. More specifically, the pediatric patients had a higher amplitude of low frequency fluctuation in the right Broca’s region and a lower amplitude of low frequency fluctuation in the fusiform gyrus bilaterally (21).

Deficits in manipulating the sounds of speech (phonological processing) suggests dominant hemisphere pathology. Spoken words (eg, the word “epilepsy”) can be broken into syllables. Syllables are groupings of a vowel sound with or without surrounding consonants (eg, “e/pi/lep/sy”), and can be further separated into basic and distinctive sound units; phonemes (eg, “e/p/i/l/e/p/s/i”). Phonological paraphasic errors occur in conversation when patients substitute phonemes in spoken words (eg, saying “manguage” for “language”). Phonological errors can also occur at a broader scale, for instance when a similar-sounding word is substituted for a target (eg, “sandwich”). Such errors are seen following impairment through a range of regions within the language system, including Broca’s and Wernicke’s areas, and the angular and supramarginal gyri.

Deficits in reading written words can also indicate dominant hemisphere pathology. In the same way that words can be broken down to the level of phonemes, written words can be broken into “orthographs.” Inferior parietal cortex, and particularly the angular and supramarginal gyri, are particularly important in mapping these written representations of language with their spoken equivalents (orthography-phonology mapping). Intraoperative stimulation of the dominant angular and supramarginal gyri impairs reading ability and results in paraphasic errors (74). Deficits are most apparent when patients are pronouncing orthographs alone or in unfamiliar groupings; therefore, it is better to test for deficits than to rely on subjective report. Deficits are best elicited through tasks requiring patients to vocalize single (eg, /a/, /f/, /th/) or multiple (eg, “qop,” “sigtrav”) orthographs.

Posterior temporo-occipital cortex, and in particular a region of the fusiform cortex known as the “visual word form area,” is also critical in the processing of word forms, and lesions in this area can result in alexia and agraphia (75; 83). Of direct relevance, epileptogenic pathology in this area can completely disrupt the ability to read and write a given language (47).

Impaired naming, or “word-finding difficulties” indicate impairment of the language-dominant hemisphere and, in their purest form, reflect damage to an area of cortex critical in associating information about an item (semantic knowledge, visual imagery) with the item’s name. Subjectively, such patients with naming difficulties report hesitation in spontaneous speech and difficulty with word retrieval in conversation. In the clinic, this manifests as hesitation in conversation and a tendency to provide (often imprecise) synonyms for words (ie, circumlocution, or “talking around the word”). For instance, instead of saying “I use the bus to get to work” a patient may alternately report “I use the… [pause] that vehicle… [pause] the big vehicle with lots of people… the bus to get to work.” Such deficits are typically less prominent in individuals with higher baseline cognitive function who can often fluently substitute semantically-related alternatives for words that can’t be retrieved; for instance, “I use the [brief hesitation] public transit to get to work.” An area of inferior-lateral temporal cortex usually called the basal temporal language area (BTLA) is critical for this process (27; 01). In a large study, it was shown that resection of this area, which falls predominantly within the fusiform gyrus, predicts around half of the variation in language (naming) decline after surgery in the dominant temporal lobe (15).

For the most common form of surgery, Busch and colleagues derived and then validated (in a separate sample) a simple nomogram for predicting naming decline after a temporal lobe resection (19). Using a worksheet provided by Busch and colleagues, the risk of moderate or greater decline can be calculated simply using (1) side of surgery, (2) age at seizure onset, (3) age at surgery, (4) education, and (5) sex.

Selective impairment of the ability to write in the presence of preserved language in a seizure disorder can indicate damage to the dominant frontal or temporo-occipital regions. Specifically, stimulation of cortex in Exner’s area, a region of the posterior middle frontal gyrus rostral to the motor strip’s hand area (29), can selectively disrupt the ability to translate orthography into written language without impairing oral language or motor movements (73). Clinically, highly focal lesions here can manifest as alexia and agraphia (02). This area can also be mapped using fMRI and direct stimulation (73; 09).

Impaired continuation of speech. Epilepsy patients with acute supplementary motor area damage, which can follow a seizure or the growth of a tumor within this region, may exhibit a dwindling aphasia. That is, sentences gradually fade shortly after they are initiated so that the person falls silent after speaking a few words. This reflects the supplementary motor area’s critical role in sequencing motor movements (71). Such impairment is typically transient as the contralateral supplementary motor area can support function lost by ipsilateral supplementary motor area damage, and the extent of damage to this area corresponds with functional deficits (22). Note that these deficits or related deficits can be enduring if callosal fibers are lesioned with the dominant supplementary motor area (28).

Impairment of Broca's area, in the posterior third of the inferior frontal gyrus, can manifest in multiple ways. At the broadest level, insult to Broca’s area can result in word-sequencing deficits such that patients produce nonfluent, halting speech that is lacking in function words (agrammatism). For instance, the sentence “The cat sat on the mat” may be produced as “cat… mat… sitting”. Alternately a patient may describe their visit to the doctor “Um… Friday, Friday… visit… Mum… I visit… doctor and ah… heart… check heart.” Patients with Broca’s aphasia may also experience difficulty comprehending sentences that involve complex syntactic structures. For instance, those with Broca’s aphasia make errors when asked a question in the passive voice with noncanonical word order (object-verb-subject), such as “The Pistons were defeated by the Celtics. Which team won?” In contrast, they are able to answer the question when the active voice is used (canonical word order: subject-verb-object), for example, “The Celtics defeated the Pistons, which team won?” (16; 63). That is, comprehension of semantics at the single-word level remains intact, but processing utterances relying on syntactic movement is lost. In line with this, Broca’s area is also engaged during silent reading, which does not involve overt speech production (18). Further, distinct regions within Broca’s area are now known to be preferentially engaged in semantic (anterior region; pars orbitalis), syntactic (medial region; pars triangularis), and phonological processing (posterior region; pars opercularis) (16; 32). Concurrently, these subregions are functionally coupled, which means that they are activated in parallel and not sequentially (23).

In contrast, damage to Wernicke’s area in the posterior component of the superior temporal gyrus impairs the selection of language content. An excellent reconceptualization of this area has highlighted the anatomy of this area, which has long been defined differently by various authors (14). The posterior superior temporal gyrus and supramarginal gyrus in the dominant hemisphere appears critical for retrieving and ordering the sounds of speech. When Wernicke’s area is damaged, all speech production tasks can be disrupted. Patients with lesions in Wernicke’s region, then, may offer responses to questions with the correct grammatical structure but incorrect content. In an error of selection, a patient may offer the word "dog" rather than “cat” (semantic paraphasia) or, if phonological processing is disturbed, “cot” or “cap” (phonological paraphasia). With extreme impairment, this may happen to the point where sentences become unintelligible (eg, “The cat sat on the mat” becomes “The animal walked on the green door”). Further, sequencing may be disturbed at a more fundamental level and interfere with phoneme selection. Here speech can becomes jargonistic; instead of “the cat sat on the mat,” the individual with aphasia may say “atat it mit mat mat at.” Such impairment is more likely to include more posterior temporal or inferior parietal pathology.

Ictal aphasia. Although all the forms of language impairment discussed above can be enduring, there may be transient phenomena that occur only during a seizure (ie, “ictally’”) (87). Depending on seizure focus, different types of language impairments have been reported during ictal aphasia. For example, seizure activity in the basal temporal regions of the left-dominant hemisphere has been associated with jargon aphasia (fluent but highly incoherent language production accompanied by disrupted language comprehension). The transient language impairments can be helpful in assessing seizure focus (84).

Pathology of the nondominant hemisphere

When pathological processes predominantly affect the nondominant hemisphere, especially posterior aspects, visuospatial function is typically affected (91). Patients may report a range of deficits on tasks requiring judgment of spatial relationships. For instance, they may misjudge the boundaries of a vehicle when parking; they may be clumsy when picking up or using items; or they may have difficulties completing hobbies that require fine visual discrimination (eg, knitting, tool use). Any such difficulties must be differentiated from primary issues with motor function or object agnosia. In the clinic, deficits may not be readily observable but can be inferred from reported behavior (eg, misjudgments of the edges of furniture and awkwardness when navigating). Difficulties in copying a line drawing can readily reveal significant visuospatial deficits, though again it is important to rule out more basic problems like difficulties in visual or motor function. On neuropsychological assessment, tasks sensitive to visuospatial deficits range in difficulty and demands--from simple judgments of line position (eg, judgment of line orientation task) (11) or the ability to reassemble the cut-up pieces of a line drawing, as in a puzzle (eg, Hooper Visual Organization Test) (35) to the ability to reason using complex visual stimuli at the most complex level (eg, matrix reasoning task of the Wechsler Adult Intelligence Scale IV) (95). An alternative to the hemispheric material-specific hypothesis, the response bias model, holds that regardless of material content, the left hemisphere is more prone to errors of omission, whereas the right is prone to errors of commission; there is evidence that both models independently correlate with lesion laterality (33).

Although language impairments are not typically reported in patients with right hemisphere epilepsy, individuals with right-lateralized focus have been found to display deficits in discourse. By discourse we understand the ability to select information that is contextually relevant and integrate it to infer the meaning correctly (44). Prior research on patients with damage to the right hemisphere suggests that this hemisphere is engaged with discourse. When the ability to process discourse is disrupted, functional communication can be impaired (79).

Memory

Material-specific memory impairment (temporal lobe epilepsy). Impairment of memory is a cardinal cognitive sign of mesial temporal lobe damage. Bihippocampal lesions are associated with a dense impairment of the ability to recall richly spatial, temporal, and personal memories (episodic memories) while the ability to learn new semantic information is preserved (90). More importantly, in presurgical patients, the fact that unilateral temporal lobe impairment is associated with impairment of memory for specific types of material (theory of material specificity) is the cornerstone of neuropsychological assessment in epilepsy (55; 82; 58). This theory has been modified somewhat with accumulating evidence (78) to reflect the statement that verbal memory performance is an index of dominant temporal lobe integrity, whereas nonverbal memory reflects more bilateral mesial temporal function.

Deficits in memory for verbally presented material can manifest as trouble recalling past conversations, material that has been read, details from the news, or movie plots. Nonverbal memory deficits often manifest in difficulties remembering one’s way around familiar routes (eg, the way to the store) or places (the family home) or in difficulties navigating a new environment (eg, a new home or suburb). In neuropsychological assessment, the central sign of dominant mesial temporal impairment is a deficit in the formation of novel, arbitrary associations (77). The ability to learn semantically unrelated word pairs is commonly assessed (eg, “apple-silver”) (Wechsler Memory Task IV: Verbal Paired Associates). Such verbal associations preferentially rely on the mesial temporal lobe, whereas recollection of semantically related word pairs (eg, “walk-run”) tends to tax the lateral temporal cortex in the dominant hemisphere (94). Other dominant memory measures that rely on lateral temporal structures include tests of memory for verbal passages, such as the Logical Memory subtest of the Wechsler Memory Scale IV.

A nonlateralizing indication of mesial temporal lobe impairment is failure to learn more with repetition. For instance, while a patient may be told a shopping list of 15 items by their partner 5 times, they may only ever be able to recall 6 of the items at most. This indicates an inability to learn beyond the capacity of one's short-term memory (approximately synonymous with “working memory”), a process more heavily dependent on the frontal lobes.

For temporal lobe resections, equations are available to predict the risk of postoperative decline on a number of key memory measures (20). Variables used in prediction include surgery side, baseline performance and, in some cases, whether the hippocampus was resected or not, and patient’s education level. The risk of decline can be calculated using the worksheets provided by Busch and colleagues (20).

Executive contributions to memory. Specific forms of memory deficit may indicate impairment of the executive functions. Impaired executive function is associated with a number of forms of brain insult but is most typically associated with damage to the frontal lobes. The executive skills are a set of functions used to plan, organize, and monitor behavior. They are engaged in memory tasks in a number of ways.

Executive dysfunction can interfere with strategy use in memory. For instance, when learning a long list of items to be bought from the supermarket, they will be better remembered if items of different types are grouped (for instance; fruit and vegetables; meats; dairy and so forth). Patients with executive dysfunction are less likely to use this strategy and may focus on a stimulus detail at the expense of broader organizational characteristics. Such patients can benefit from cueing, which orients them to the organizing principles of the material. These patients may perform poorly on spontaneous recall but may dramatically improve when cued (ie, asked what meats, vegetables, and dairy items they needed to get). These category cues provide explicit structure, so performance improves once the executive requirements of the task are diminished.

Patients with executive difficulties may also find it hard to distinguish learned items from new items, particularly when these items share features. This is because during the retrieval of information it is necessary not only to activate representations of stored material but also to inhibit or suppress other salient but unrelated material. Patients with executive deficits can have difficulties engaging inhibition. For instance, if an individual who needed to collect butter, milk, and cheese from the supermarket is later asked if they needed to collect these 3 items, they may say yes to each of them, but when then asked if they also needed to eggs and yogurt assent to these items as well. On formal neuropsychological testing, this is assessed by tests of recognition where, after a person has had an opportunity to recall information they attempted to learn earlier, they are given a list of both old, learned items and new items and asked which they encountered previously. The relationship between the number of endorsed items that were (“hits”) and were not (“false alarms”) presented gives an index of discriminability, which can indicate executive deficits. Patients with executive dysfunction may have higher rates of “false alarms” than expected.

Attending to and organizing incoming information is another frontal lobe function that if impaired can impact memory. This is particularly noticeable when the incoming information exceeds working memory capacity; in these cases, tests that include repetitions of stimuli may be more easily performed than those in which the information is heard just once. Patients with frontal lobe impairments may fail to engage efficient strategies such as chunking or organizing semantically, relying instead on inefficient serial learning strategies.

Other markers of frontal lobe impacts on memory include interference effects. Examples of interference are worse performance on a second word list compared to the first trial of a previously presented word list; an early plateau or inverted learning curve, where multiple attempts to retrieve from the same list result in increased difficult with recall; and release from proactive interference occurs when there is a break in time or a distraction that results in better performance. At times patients with proactive interference can perform as well or better after a delay compared to immediate recall, particularly on word lists with multiple presentations.

It is also possible that individuals with more severe executive difficulties may confabulate when relying on memory. Here, families may note that when the patient is asked about daily events they “make things up.” Such patients typically take key elements of events and combine these with other information. For instance, when asked what items they were going to get from the supermarket, the patient may respond “milk, cheese, eggs... And then I had to go to the bakery to get bread and some cakes, because you like cakes, and after that I had to go to the drug store...”

Cognitive phenotypes and ongoing research. Most of our understanding of the relationship between neural impairment and cognition in epilepsy has proceeded, historically, from our general knowledge of brain-behavior relationships, case studies, and data from small samples. This has recently been replaced by a move towards multi-site studies with many hundreds of patients that can identify cognitive phenotypes (70) or derive and then independently validate patient-specific algorithms for predicting cognitive decline (19; 20). One notable recent contribution re-examined the 3 core cognitive profiles previously shown to exist in temporal lobe epilepsy, which is the most common form of epilepsy (70). These profiles were again seen in a sample of over 400 patients and included (1) generalized impairment across cognitive domains, (2) impairment of memory and language, and (3) no cognitive impairment. When clinically-usable criteria were used, these profiles were identified in approximately one third of patients.

Clinical variables modifying the cognitive presentation

A number of variables critically determine the nature of cognitive impairment seen in epilepsy. One of these is the age at which seizures commenced. When seizure onset occurs early in life, as a general rule, it is more likely that cognitive functions, eg, localization of language and verbal memory to the left hemisphere, will be disturbed.

The precise age at which neural insult occurs is also important when considered in light of a child's neural, cognitive, and psychosocial development. When seizures commence at a critical developmental point (for instance, during language development), the course of development can be disturbed; function acquisition may be delayed, occur atypically, or potentially be discontinued all together (25). A study investigated a vast range of pediatric epilepsy syndromes (eg, temporal lobe epilepsy, absence seizures, juvenile myoclonic epilepsy) (37). The authors reported a high degree of common language impairments (eg, poor vocabulary performance) shared across the epileptic groups when compared to healthy controls. Deficits may not initially be apparent but may appear as children subsequently fail to acquire expected functions (“growing into deficits”). It is also important to note that insults to the brain actually have a multiplicative impact, not simply a cumulative one (“double hazard” model) (03). That is, consecutive negative events tend to have disproportionately greater impact on an individual. The effects of a head injury, ADHD, or chronic malnutrition will disproportionately influence a child who is already suffering a seizure disorder.

The age of onset is closely linked to the duration of time during which regular seizures have been occurring. The longer seizures continue, the greater the impact on cognition (31). For example, earlier onset of epilepsy and longer seizure duration has been associated with disrupted functional connectivity within the temporal lobes, which might explain deficits in naming in individuals with temporal lobe epilepsy (85). Other key seizure variables are the number of lifetime generalized tonic-clonic seizures, a larger load of antiepileptic drugs, and the number of episodes of uncontrolled seizures (status epilepticus) (31). More episodes of longer duration indicate significantly worse cognition.

Another variable affecting cognitive impairments in individuals with epilepsy is seizure location. For instance, children with frontal lobe epilepsy have been shown to have more problems with executive function as well as more widespread neuropsychological deficits than children with temporal lobe epilepsy. Compared to healthy controls, children with temporal lobe epilepsy had poorer performance on verbal knowledge tasks, tasks evaluating inhibition, and shifting in daily life (89).

Antiepileptic medications. Antiepileptic medications can be associated with significant cognitive side effects. Those side effects are not only subjective but also objectively quantifiable (65), although individuals’ responses to these medications can differ.

Perhaps most notable among these are older antiseizure medications (ASMs; also known as anti-epileptic drugs or AEDs). These include sodium valproate/valproic acid, phenytoin, carbamazepine, and phenobarbital in particular, all of which can impact attention, processing speed, and memory retrieval. More recent medications (eg, felbamate, tiagabine, levetiracetam, lamotrigine, gabapentin, vigabatrin, oxcarbazepine, zonisamide) appear to have less significant cognitive effects. An exception is topiramate, which can cause significant impairment of memory, reaction speed, and language (naming), as shown in a randomized, double-blind study showed (46). These effects are marked at higher doses and can be ameliorated with good titration. Another study showed that use of topiramate is associated with up to 50% worse performance in fluency than gabapentin and lamotrigine (48). Phenytoin and carbamazepine have been associated with impairments in verbal memory recall (Montamedi and Meador 2004). Cognitive side effects of antiepileptic drugs are related to patient compliance and may be associated with intolerability of the drug and ultimately discontinuation. One study found up to 11% of patients on topiramate described intolerability of cognitive side effects and ultimate discontinuation of the drug, with phenytoin and zonisamide a close second and third (04). Recent fMRI studies indicate that both topiramate and zonisamide altered functional connectivity in networks essential for higher cognitive processes including language (92). Cognitive effects should be carefully considered when evaluating patients with epilepsy, and consideration of whether or not an individual could discontinue a particular medication prior to neuropsychological evaluation is warranted. Finally, subjective and objective cognitive deficits are more pronounced in individuals treated with multiple antileptic drugs (polytherapy) than in those treated with a single antileptic drug (monotherapy) (65).

Psychiatric comorbidities and cognition. Psychological factors may have very significant effects on cognition and can alter the presentation of localizing cognitive impairments. Symptoms of anxiety and depression result in reduced attention, working memory, and processing speed in addition to other deficits (24). Patients with epilepsy have very high rates of both anxiety and depression (93); these symptoms may in turn affect their cognitive profiles; for example, finding memory impairments that may suggest localized, mesial temporal deficits, when in fact the impairments may result from poor encoding due to reduced attention and concentration secondary to depression or anxiety. One study found that patients with early seizure onset had impaired autobiographic memory that was linked to young onset of epilepsy, reduced working memory, and more frequent seizures (69). In contrast, patients with late seizure onset also had deficits in their autobiographic memory but those were associated with depression and lesion presence (as identified by MRI) (69). For these reasons, neuropsychological testing should include comprehensive psychological evaluation and consideration of psychiatric comorbidities.

Clinical vignette

Patient G was a 32-year-old, right-handed woman with 18 years of education, under consideration for a left temporal lobectomy. She had epilepsy since the age of 9 and the seizure frequency continued until her current rate of 2 to 6 seizures every night. She was referred for neuropsychological assessment to determine whether her neuropsychological profile was localizing, consistent with other diagnostic tests, and to characterize the risk of cognitive dysfunction following surgery.

Medical history. Ms. G was a 32-year-old right-handed female with a history of nocturnal seizures for which she was considering surgical evaluations. She had normal growth and development with no other significant health issues and no known seizure risk factors, including no febrile convulsions, head trauma, or other neurologic events.

Seizure history. Seizures began at the age of 9 and have always occurred nocturnally. She was initially treated with carbamazepine and was seizure free with the exception of rare events associated with missing medications, but began having breakthrough seizures at age 31. Several attempts to change to other medications resulted in seizures, and she remains on carbamazepine with the recent addition of Onfi.

Seizure semiology. Seizures often occur between 2 AM to 3 AM, and are described as an arousal with a feeling of “electric” sensation in her right face, sometimes followed by R facial twitching and RUE uncontrolled movements or scratching her face for 5 to 15 seconds. Sometimes, she would have the right facial tingling without progression. Speech is impaired during seizures and at most will have vocalization. She maintains consciousness and has never generalized. These are exclusively out of sleep (can happen in daytime if she naps). Seizures are followed by postictal fatigue. She noted seizures can cluster up to 6 to 7 in 1 night.

Imaging. MRI revealed no frank structural abnormalities and no evidence for mesial temporal sclerosis; there was a suggestion of an abnormality in the left paracentral/cingulate border which might represent a cortical dysplasia type 2. FDG-PET imaging was negative.

Functional MRI. FMRI was conducted in English and Spanish. FMRI showed bilateral language activation for both English and Spanish.

Magnetoencephalography. MEG results showed modeled dipoles corresponding to spike activity localized to the posterior aspect of the left inferior frontal gyrus/frontal operculum clustering in the region of the pars opercularis.

EEG. Standard awake scalp EEGs were normal. Videotelemetry was conducted as an in-patient. A total of 6 seizures were captured during her stay, representative of her typical seizures at home. The evaluation indicated that “although the seizures had stereotyped onsets, the electrographic onsets for these events were not evident on scalp EEG.”

Psychosocial history. Ms. G was raised with both parents and an older sibling in South America, emigrating to the U.S. at the age of 5. She is trilingual, with English and Spanish learned simultaneously as second languages; she speaks English fluently and is most comfortable with English as her primary language. She has no history of trauma or abuse and no psychiatric or substance abuse history.

Educational and employment. Upon emigration to the U.S. the patient repeated a grade and was in ESL classes, but then moved to a gifted and talented program in the public school system. She has a Bachelor’s and a Master’s degree from highly competitive schools and works full time in her profession in health care.

Psychological risk. The patient does not report significant psychological concerns and is not significantly depressed or anxious. She lives with a significant other and will spend some nights with her parents.

Presenting complaints. Ms. G reported when she was taking Onfi; her cognition and processing speed were very poor. Since taking it at night only, she reported some improvement, but is not back to baseline. She noted difficulty with both written and verbal expression, occurring more frequently towards the end of the day. She complained of problems in word retrieval in 1 language but not another (eg, English but not Spanish), and noticed it happening more when speaking English which she uses about 75% of the time. She also noted problems multitasking, but this improved on Onfi. She stated that long-term memory for events before 5-years-old was difficult, but denied problems with recent events with use of compensatory strategies (eg, taking notes). She denied any difficulties with planning, organizing, spelling, and reading comprehension. She reported increased emotional lability on Onfi and some depressed mood due to uncertainty regarding seizure control and lack of sleep; this did not last for more than 1 to 2 days.

Neurocognitive results. No factors compromising task validity were apparent. On the basis of her educational and employment history, premorbid function was estimated to fall within high average range.

The patient’s intellectual abilities were found to be at the high end of the average range, and high average in verbal comprehension index. Verbal reasoning skills (similarities on the WAIS-IV) were high average, and visual spatial reasoning was at the upper end of the average range (66th percentile).

Language skills on formal testing were intact. She made rare paraphasic errors in spontaneous speech and 1 pronunciation error in repetition. Otherwise, fluency, naming, verbal abstraction, word knowledge, and reading were all in the normal range.

Visual skills including visual organization, spatial problem solving, and face recognition were all intact and above the 50th percentile. In contrast, her copy of a complex figure was careless and poorly organized, and below the 1st percentile.

Verbal memory scores were mixed and are discussed in detail in the formulation section below. Her performance on verbal paired associates was average both immediately and after a delay with 91% retention. On serial list learning (California Verbal Learning test) she showed first trial performance in the borderline range, an early plateau in her learning curving, overall in the low average range, but 90% retention of material initially recalled after a delay. Recognition was 100% with 2 false positives; her learning style showed a reliance on serial order rather than semantic processing. On a passage task where the information was presented only once, she was in the 27th percentile at immediate recall and 7th percentile following a delay.

Visual memory for a complex figure showed borderline performance on immediate recall and no loss of information following a delay. Intermediate difficulty material (visual reproduction, WMS) was high average on immediate recall but borderline after a delay. Given visual material with 3 repetitions (Briev Visual Memory Test), performance was average with 100% retention after delay.

For executive functions, word fluency, response inhibition, and switching were all in the high average range, whereas inhibition with switching (DKEFS Trails) was in the low average range. Concept formation and alternation was intact.

Measures of anxiety and depression showed minimal concerns, and the quality of life scores were all within normal limits.

Case formulation. This patient was bright and high functioning with good work and education history. Seizures have significantly impacted her life in the following ways:

Although most cognitive performance was at average or above average range, several areas stood out as difficult. Of note, these were not limited to a single modality. At first glance, the presence of significant verbal and visual memory deficits may suggest a mesial temporal lobe onset. Verbal memory deficits are linked to left temporal lobe dysfunction and are common in left temporal lobe epilepsy. However, a closer examination reveals an atypical pattern. First, the patient’s learning curve on list learning shows reduced first trial performance on a word list (5), which is below that expected for someone of her educational attainment. The learning curve increases but plateaus by the third trial, and the patient actually performs worse at 1 later trial than an earlier trial (5, 9, 11, 12, 11). The interference word list shows a raw score of 4 items less than the first list and far below her baseline expectation. Each of these findings suggests proactive interference, a more frontal lobe finding. In comparison to short delay recall of a word list, the long delay showed no appreciable decline. However, there were multiple intrusion errors and repetitions, both frontal lobe signs. Together, the word learning test does indicate verbal memory problems, but the origin appears more frontal in nature rather than temporal. On paragraph learning, there is a significant delayed recall deficit. Verbal paired associates learning was average and there was no significant information loss after a delay. Initial encoding of large amounts or complex material always produced reduced performance. There was evidence for difficulty organizing incoming visual and verbal information. A close examination of the verbal list learning task showed that the patient did not organize information by semantic category, choosing instead an inefficient, serial learning approach. With the exception of the passage reading task, she demonstrated no significant loss of information after a delay. In this task, information is not repeated, placing a larger burden on attention and organization skills that are frontally mediated. Further, the patient made multiple intrusion and repetition errors during memory performance. Thus, although the patient had significantly lower memory performance than expected, particularly verbal memory, the specific nature and pattern of memory problems was more consistent with a frontal, rather than a mesial temporal, onset. This formulation turned out to be in agreement with the videotelemetry, which ultimately found evidence for central-frontal EEG abnormalities.

Surgical considerations. Because the cognitive profile, MRI, MEG, and EEG were not consistent with a standard mesial temporal lobe seizure onset, the patient was recommended for additional diagnostic evaluation, including FMRI, Wada testing, and potentially depth electrodes. Given her extratemporal onset, lack of strongly localizing imaging and EEG findings, high educational and occupational achievement, and overall high level of functioning, surgery was considered more likely to produce postoperative cognitive deficits in this patient that could potentially have a significant impact on her quality of life. Wada testing was recommended to better determine which hemisphere was essential for language. Further evidence for seizure localization, perhaps with depth electrodes, will be considered. This difficult case demonstrates the important role for careful evaluation of neuropsychological findings that appreciates the complexity of language and memory organization in the human brain, and the need to interpret findings with careful consideration of not just whether cognition is impaired, but exactly which underlying processes contribute to that impairment and how these processes are represented neuroanatomically.

Diagnostic workup

Neuropsychological assessment. A primary goal of preoperative neuropsychological evaluation is to identify potential discrepancies in cognitive deficits and seizure localization results including EEG, MEG, and MRI. Neuropsychological assessment ideally starts from a standardized battery then is tailored to the patient’s clinical complaint and the referral question. It will typically include an assessment of premorbid cognitive function against which impairment can be identified (eg, the Test of Premorbid Function; TOPF). An assessment of global intellectual function is then necessary (eg, Wecshler Adult Intelligence Scale IV [WAIS] or Wechsler Abbreviated Intelligence Scale II). This allows identification of global impairment, which may occur with significant GTCS or status epilepticus, or with a long duration of regular seizure occurrence. Components of the WAIS provide measures of attention and working memory (digit span) and processing speed (coding). Again, impairment of these is most likely to detect diffuse cognitive pathology. Assessment of language functioning should include a range of tasks to assess the different skills detailed above. This may include assessment of naming ability using both visually (eg, Boston Naming Task II) and auditorily presented stimuli (eg, Boston Diagnostic Aphasia Inventory [BDAE] 3 Auditory responsive naming). The BDAE also contains a series of measures examining reading and auditory comprehension, expression, and repetition. The Woodcock Johnson III contains a particularly nice measure of orthography-phonology mapping (“Word Attack”) sensitive to angular or supramarginal impairment. Assessment of spatial function may include basic tasks in which line orientation is judged (Judgment of Line Orientation task) and visuoconstructional tasks (eg, Hooper Visual Organization Test). Memory assessment should include both verbal and nonverbal measures per those listed above (Wechsler Memory Scale IV: Verbal Paired Associates, Logical Memory & Visual Reproduction; Rey Complex Figure; California Verbal Learning Task II).

Psychosocial and psychological assessment. Psychological and psychosocial assessment is a critical part of presurgical evaluation typically performed by neuropsychology. A significant proportion of epilepsy patients have comorbid psychological conditions; further, psychiatric. Depression is found is 20% to 55% of patients with epilepsy and is the most frequent comorbid psychiatric diagnosis in epilepsy (88). Notably, depression is higher in patients with mesial temporal sclerosis than other types of epilepsy (64). Thirty-two percent of all deaths in epilepsy are from suicide (61), highlighting the importance of making a comprehensive risk assessment presurgically as well as underscoring the need for both presurgical and postsurgical management of psychiatric symptomatology. Psychosis can occur postictally; rarely, surgery can result in worsening of psychiatric symptoms or de novo psychosis. These outcomes are more common in patients with more bilateral abnormalities, a smaller contralateral amygdala, and seizure onset outside of the mesial temporal lobe (80). Mental health care for individuals with epilepsy remains challenging due to the lack of standardized procedures and integrated mental health specialists. In a study, over 60% of patients with epilepsy suffering from a psychiatric comorbidity were not receiving care for their mental health (54).

Quality of life and surgical outcome. A commonly used standardized instrument to assess real-life impact of epilepsy and surgical outcome is the Quality of Life in Epilepsy (QOLIE) (59). There is strong evidence that appreciation of the likely cognitive outcomes following surgery result in superior outcomes, in part due to having realistic expectations of the potential changes is cognition postoperatively. Careful assessment of potential areas of cognitive decline can help plan postoperative interventions and set up realistic expectations for quality of life.

Invasive assessment. More direct mapping of cognitive function is essential to evaluate the assumptions underlying the assessment (namely that the left hemisphere represents language) and to quantify the likelihood of postsurgical impairment in memory and language.

fMRI is an invaluable noninvasive method for mapping cognitive function presurgery. It is almost certain that it will eventually replace the Wada as the method of choice for establishing the laterality of verbal memory function. A number of authors have argued that clinical fMRI has now matured to the point that it can replace Wada testing for the purposes of language lateralization (13). Indeed, many centers have replaced the Wada test with fMRI for language lateralization. However, using fMRI for language lateralization in epilepsy remains an issue of ongoing debate, especially given the consequences of incorrectly localizing language (50). For instance, a study showed that in a series of 229 consecutive patients at a comprehensive epilepsy center, language lateralization by fMRI and Wada was discrepant in 42 instances (38). In the majority of these cases (38), a measure involved a determination of bilateral language function; in a further 4 isolated instances, Wada indicated left lateralization where fMRI indicated right (1.75%). There were unfortunately no significant predictors of discordance. Moreover, they subsequently showed that when fMRI and Wada testing give differing judgements of language dominance, fMRI more accurately predicts language decline (38). Although this demonstrates fMRI can replace Wada testing for language lateralization, it is notable that this is only true in the presence of a highly standardized and reproducible form of language fMRI, and that many epilepsy programs likely do not currently have access to this (10). Although it is not currently feasible, our ability to localize specific regions within the dominant (and nondominant) hemisphere is an area where fMRI will be particularly valuable, and efforts are currently under way to validate it for this purpose.

With respect to memory, there is currently no widely accepted, clinically validated protocol for lateralizing function, though there is a growing literature on memory activation paradigms in epilepsy (67; 81). Magnetoencephalography (MEG), though not as widely available as fMRI, appears to have a similar ability to lateralize language functions (96).

Direct electrocortical stimulation mapping (ESM) with superficial grids or depth electrodes (phase 2) is typically completed if it is highly likely an epileptogenic focus in eloquent cortex. This involves a craniotomy followed by installation of one or more grids of electrodes and/or depth electrodes over the relevant cortex before the craniotomy is closed and the patient returns to ward. A neuropsychologist then evaluates cognitive function with tasks appropriate for the relevant region (eg, naming, reading, etc.) while an electrophysiologist systematically stimulates different electrodes through the grid. The results are then correlated with the precise location of the electrodes as determined by imaging. This can be distressing for patients--the occurrence of 1 or more seizures is virtually inevitable--but can be an invaluable means of determining whether a resectable seizure focus exists. In the past decade, there has been a growing interest in adding stereo-electroencephalography with electrocortical stimulation mapping, a technique that can help target sulci and deep brain structures, such as the hippocampus or insula (76; 05). There has also been some recent focus on the fact that, in electrocortical stimulation mapping, considering neural activity both distal and remote to the site of stimulation through the variation of polarity during stimulation can improve our understanding of seizure propagation and, potentially, the clinical effects of resection (08).

If grid mapping is not feasible, awake intraoperative mapping provides one final option for mapping cognition. Here the patient is awakened in surgery following craniotomy, and cognitive function is directly evaluated as in grid mapping but with the neurosurgeon directly stimulating cortex either prior to or during resection. Again, this can be an invaluable method so long as it is tolerable to the patient but is frequently complicated by epileptic activity during stimulation. Nonetheless, direct cortical stimulation remains the gold standard to determine the location of brain regions critical for cognitive functions, particularly language (34).