Carl W. Bazil MD, PhD*
Timothy A. Pedley MD**
*Assistant Professor, Department of Neurology, Columbia University, New York, New York; and Director, Clinical Anticonvulsant Drug Trials, Comprehensive Epilepsy Center, New York Presbyterian Hospital, New York, New York
**Professor of Neurology, Department of Neurology, Columbia University, New York; and Neurologist-in-Chief, the Neurological Institute of New York, New York Presbyterian Hospital, New York, New York
Twentieth-century advances in experimental and clinical neurophysiology and in pharmacotherapeutics have revolutionized the diagnosis and treatment of epilepsy. Electroencephalography (EEG) and antiepileptic drugs (AEDs) are inextricably linked in day-to-day practice. Many AEDs have effects on routine EEG and evoked potential recordings, and familiarity with these effects is necessary for accurate test interpretation. At the same time, EEG findings assist in making decisions regarding initiation and withdrawal of AED therapy and, in some cases (e.g., infantile spasms, absence seizures), in determining if the desired therapeutic effect has been achieved.
In this chapter, we review common effects of AEDs on background and interictal EEG activity and on evoked potentials, and we discuss the role of EEG in therapeutic decisions.
ELECTROENCEPHALOGRAPHIC EFFECTS OF ANTIEPILEPTIC DRUGS
The most clinically relevant effects of AEDs are on EEG background activity and on epileptiform discharges. Because of some notable differences, the effects of major agents or classes of agents are considered separately.
Visual Analysis of Electroencephalographic Background Activity
At therapeutic doses, all benzodiazepines increase the amount and prominence of EEG rhythmic fast activity, usually in the 25- to 35-Hz range (1, 2, 3, 4). This effect may be first noticed or most obvious during drowsiness. The voltage of the alpha rhythm usually is correspondingly reduced (4). To some extent, these effects are age and time dependent: EEG changes are more dramatic in younger patients than in older ones, and acute administration results in more prominent beta activity than does chronic use. With a few benzodiazepines (e.g., clonazepam), EEG effects can be dose related (5).
EEG changes related to benzodiazepine intoxication occur sequentially. The rhythmic beta activity increases in voltage and becomes more sustained. This is followed by increasing amounts of slow-frequency activity, which parallel depression of cognitive abilities and increasing lethargy. With onset of coma, the mixture of spindle-like, fast-frequency activity superimposed on continuous slow theta and delta activity resembles EEG patterns seen with barbiturate overdose (6,7).
At therapeutic doses, barbiturates also entrain faster frequencies into rhythmic patterns, usually in the 18- to 25-Hz range, which are most prominent frontally (6,8,9). As the dosage is raised further and clinical toxicity appears, there is a progression of increasing slow-frequency activity combined with nearly continuous, waxing and waning spindle-like rhythms in the 4- to 12-Hz range (“barbiturate spindles”). Biphasic or triphasic, sharp, slow delta waves may occur (10). Consciousness usually is lost by the time rhythmic or near-rhythmic delta activity predominates with superimposed barbiturate spindles (11). Deep coma is accompanied by a burst-suppression pattern or even electrocerebral inactivity (6,11). This sequence of changes is illustrated in Figure 2.1. With early medical intervention and intensive support, both clinical and EEG features of barbiturate overdose are completely reversible. As the drug is cleared from the system, a reverse sequence of changes is seen (9).
FIGURE 2.1. Effect of increasing concentrations of intravenous thiopental on the electroencephalogram. A: Increased fast-frequency activity. B: Barbiturate spindles of 7 to 10 Hz. C: Generalized slow-frequency activity. D: Burst-suppression pattern. (From Clark DL, Rosner BS. Neurophysiologic effects of general anesthetics: I. electroencephalogram and sensory evoked responses in man. Anesthesiology 1973;38:562-582, with permission.)
Rhythmic fast activity is so typical of acute barbiturate and benzodiazepine administration that its complete absence suggests diffuse underlying organic brain disease (e.g., severe static encephalopathy). Focal or regional asymmetries indicate localized parenchymal dysfunction, which may be due to a chronic epileptogenic focus (12,13) or a cerebral lesion such as porencephalic cyst, tumor, or infarction (6,14). Emergence of EEG fast activity with diazepam has been associated with a good prognosis for seizure control in patients with epilepsy (15).
Phenytoin usually has no visually discernible effect on the EEG at therapeutic doses, regardless of route of administration (16). No changes were detected during 4 hours of continuous EEG recording in four patients with epilepsy given a 10-mg/kg intravenous (i.v.) dose of phenytoin, even though all patients showed clinical signs of toxicity (17).
At plasma concentrations above 20 µg/mL, phenytoin can slow the mean alpha rhythm frequency slightly, although this is not a consistent effect until there is clinical evidence of drug toxicity (16). This effect may also occur in patients receiving as little as 250 mg of phenytoin i.v. (18). More pronounced symptoms and signs of neurotoxicity are accompanied by progressive EEG changes: increased theta range activity (17), intermittent rhythmic delta activity, and, with severe toxicity (plasma levels >45 µg/mL), high-voltage delta activity (16) (Figure 2.2).
Carbamazepine can produce several mild effects on the EEG; these are inconsistent and usually of no clinical consequence. At therapeutic drug concentrations, the most commonly reported changes are mild slowing of the mean alpha rhythm frequency and increased amounts of random theta activity (19,20). Such changes are much more pronounced with toxic doses (6). In a placebo-controlled, double-blind
trial involving 40 patients with epilepsy, carbamazepine was said to increase theta activity and decrease alpha activity significantly compared with placebo (20), but treated patients also were receiving phenobarbital and phenytoin. Similar EEG effects have been reported in psychiatric patients treated with carbamazepine (21), as well as in patients with epilepsy compared with phenytoin using a randomized, crossover design (22).
FIGURE 2.2. High-voltage delta activity in a patient with severe phenytoin toxicity. (From Roseman E. Dilantin toxicity: a clinical and electroencephalographic study. Neurology1961;11:912-921, with permission.)
Several studies of valproate have found no effect on routinely acquired EEG background activity at therapeutic plasma concentrations (23, 24, 25). Villarreal et al. (25) studied 25 patients with absence seizures, most of whom had other seizure types and remained on AED polytherapy during the study. Analysis of four channels of EEG recorded using a telemetry system did not detect changes in background activity. Bruni et al. (24) investigated 22 patients with absence seizures before and after 1 year of valproate therapy. Visual review of 6-hour, 16-channel EEG recordings did not reveal changes in background activity. In contrast, other investigators have reported either slower mean alpha rhythm frequencies (26,27) or increased random theta activity (28), but these effects usually are accompanied by symptoms and signs of clinical toxicity (23).
Much less is known about the EEG effects of the newer AEDs, which include felbamate, gabapentin, lamotrigine, topiramate, tiagabine, levetiracetam, oxcarbazepine, zonisamide, and vigabatrin. In general, because clinical neurotoxicity is less likely to occur with most of these agents, effects on background EEG activity also are less commonly encountered in clinical practice. Only tiagabine has been studied in a published trial; it had no effect on background EEG activity at therapeutic dosages (29). For other drugs, clinical experience and a few published abstracts (30, 31, 32) suggest little or no effect on EEG background activity at therapeutic dosages.
Computer-Assisted Analysis of Background Activity
Additional information about the effects of AEDs on EEG background activity has been obtained by using computer-assisted and topographic mapping techniques (33, 34, 35). Dumermuth et al. (5) reported a correlation between clonazepam or diazepam blood levels and frontal “beta density.” They also found that these drugs decreased the power spectra of frequencies less than 18 Hz. Such quantitative measures are consistent with changes previously described from routine visual inspection of the EEG. Similarly, acute administration of phenobarbital to normal subjects resulted in increased power spectra for frequencies greater than 16 Hz (35), an effect that also would have been predicted from changes described in the routine EEG. Other studies, however, have not found any consistent difference in power spectra between untreated patients with epilepsy and patients treated with either phenobarbital (33) or carbamazepine (33,36).
Beta asymmetries after administration of i.v. thiopental (37) or diazepam (38) have been quantified using spectral analysis.
Patients treated with valproate had less theta activity and less 13- to 19.8-Hz activity than did untreated patients with epilepsy (33). A subgroup analysis showed this to be true only for patients with generalized tonic-clonic seizures; patients with partial seizures showed less delta activity in the valproate-treated group. These results are hard to understand in light of studies of routine EEG that have shown either no effect of valproate on background activity or mild increases in slow-frequency activity. If decreased delta activity were due to suppression of spike-wave discharges by valproate (25,39), this effect should have been seen to an even
greater degree in patients with generalized tonic-clonic seizures, which was not the case.
Quantitative effects of phenytoin on EEG activity have been studied in normal volunteers. Oral administration of 100 to 1,000 mg resulted in decreased power in slow-frequency bands and increased power in fast-frequency bands at plasma concentrations above 8 µg/mL (40). The magnitude of the EEG changes paralleled plasma concentrations.
A study of six patients with epilepsy concluded that phenytoin and carbamazepine increase theta and delta power slightly and slow the alpha rhythm when individual patients were compared on or off chronic drug treatment, but these changes varied widely between patients (41). Such effects may represent a physiologic marker of a patient's unique response to pharmacologic cerebral toxicity. It remains to be shown, however, whether these EEG changes correlate with objective measures of impaired cognitive abilities.
Lamotrigine does not affect EEG background activity studied by computerized methods before and after treatment (42), although another quantitative EEG study showed increased beta and decreased theta with lamotrigine, as well as increased alpha and beta frequencies with vigabatrin, and increased beta and theta activities with topiramate (43). Most study patients were comedicated with at least two other agents, and these results therefore must be interpreted cautiously.
EFFECTS OF ANTIEPILEPTIC DRUGS ON EPILEPTIFORM ACTIVITY
With a few notable exceptions, the effect of AEDs on interictal epileptiform activity is variable and inconsistent; routine visual assessment of interictal activity usually does not correlate well with seizure control. In a provocative study, however, Frost et al. (44) offered preliminary evidence that computer-derived measures of interictal spike morphology can be quantified and correlated with seizure control. In a study of 13 children with EEG spike foci and partial seizures, phenobarbital or carbamazepine efficacy was related to decreased spike voltage, decreased spike duration, an increase in the normalized sharpness of the spike, and a decrease in the “composite spike parameter,” a computed variable that expressed the relationship among these other basic waveform measures. We are unaware of further studies of this intriguing approach.
So and Gotman (45) studied the effects of high and low AED levels on patterns of ictal discharge in 56 patients with chronically implanted depth electrodes. Reduced drug levels increased seizure frequency and the number of seizures that secondarily generalized. However, low drug levels did not affect the morphology of initial ictal events, duration to contralateral spread, or coherence between discharges.
FIGURE 2.3. Reduction in interictal spike count in a group of six epileptic patients after intravenous administration of clonazepam (CZP, 0.5 mg), lorazepam (LZP, 2 mg), diazepam (DZP, 5 mg), or saline as a percentage of preinjection spike counts. (Adapted from Ahmad S, Perucca E, Richens A. The effects of furosemide, mexiletine, (+)propranolol and three benzodiazepine drugs on interictal spike discharges in the electroencephalogram of epileptic patients. Br J Clin Pharmacol 1977;4: 683-688, with permission).
The effects of AEDs on EEG epileptiform activity are discussed in the following sections for each major agent or class of agents.
Benzodiazepines are potent AEDs when administered acutely and act rapidly to suppress interictal and ictal EEG discharges. This effect has been demonstrated on both generalized spike-wave activity and focal spike discharges using rectal (46) or i.v. (47, 48, 49) diazepam as well as oral or i.v. clonazepam (1,3,47,49). A single i.v. dose of diazepam, 5 mg, or clonazepam, 0.5 mg, administered to 14 epileptic patients in a double-blind, placebo-controlled trial reduced spike discharges to approximately one-third of the baseline rate (diazepam 32% ± 12%; clonazepam 34% ± 11%). A similar but more delayed response was seen with lorazepam (47) (Figure 2.3). Intravenous diazepam aborts or attenuates photoparoxysmal responses (48,50).
Benzodiazepines have been used to help localize the epileptogenic brain region. Intravenous diazepam (50,51) or clonazepam (1) may suppress bilateral spread of epileptiform discharges (secondary bilateral synchrony) without eliminating activity at the focus itself.
Phenobarbital can reduce the amount of interictal epileptiform activity (52) and sometimes even abolish it completely (53,54). Kellaway et al. (54) monitored 12 children with partial seizures for 24 or 36 hours before and after phenobarbital treatment using computer-assisted visual analysis. In 6 of 11 patients with complete seizure control, no interictal spikes or sharp waves were seen in the posttreatment study.
In the other successfully treated patients, the number of interictal spikes decreased by 20% to 30% after treatment.
In another study, Buchtal et al. (52) found that phenobarbital, at a mean plasma concentration of 10 µg/mL (range, 3 to 22 µg/mL), reduced interictal epileptiform activity by 90% in 11 adult patients with generalized seizures who were selected because of their high rate of spontaneous interictal discharges. Some tolerance to this effect developed, because higher concentrations of phenobarbital were required to produce a similar degree of spike suppression after treatment had been withdrawn for 1 to 2 weeks. A single patient with petit mal epilepsy showed no improvement either clinically or in terms of EEG paroxysms.
Barbiturates can be used to distinguish primary from secondary bilateral synchrony. Lombroso and Erba (55) studied 82 patients with generalized seizures and bilateral spike-wave discharges. In 30 of these patients, administration of i.v. thiopental reduced diffuse spike-wave activity and allowed identification of an EEG focus. This effect was most pronounced in patients with clinical evidence of focal cerebral lesions.
Predictable EEG changes occur during barbiturate withdrawal. Drug-addicted patients showed high-voltage, bisynchronous spike-wave discharges and diffuse slowing during withdrawal (8,56) (Figure 2.4). In patients with partial epilepsy whose barbiturates are discontinued for intensive video/EEG monitoring, withdrawal may be accompanied by generalized epileptiform activity or, rarely, by the apparent appearance of new or additional loci of seizure onset (57). The clinical significance of these newly recognized “foci” is uncertain and must be interpreted in light of other clinical data.
FIGURE 2.4. Electroencephalographic (EEG) changes in acute barbiturate withdrawal. After 52 weeks of treatment with secobarbital, 600 mg daily, the drug was stopped abruptly. These bilaterally synchronous spikewave discharges appeared within 26 hours and persisted 72 hours after drug withdrawal. The EEG returned to baseline 1 week after withdrawal. (Adapted from Wikler A, Fraser F, Isbell H, et al. Electroencephalograms during cycles of addiction to barbiturates in man. Electroencephalogr Clin Neurophysiol1955;7:1-13, with permission.)
Most investigators have found that phenytoin treatment does not affect interictal epileptiform discharges (22,58). There are a few reports to the contrary, however. Carrie (59) performed nine overnight EEG studies in a single patient and reported that epileptiform sharp waves increased as the phenytoin level rose and the number of clinical seizures diminished. He speculated that increased interictal activity was a consequence of phenytoin's antiepileptic effect, which had caused “fractionation” of convulsive discharges into interictal abnormalities. On the other hand, two studies have found that epileptiform activity decreased with phenytoin treatment. Buchtal et al. (60) observed that 14 of 27 outpatients and 11 of 12 inpatients with grand mal seizures had reduced amounts of epileptiform activity with phenytoin levels over 10 µg/mL. However, approximately one-sixth of the patients had increased amounts of interictal abnormalities. Interpretation of these results was further confounded by concurrent administration of phenobarbital in some patients. Wilkus and Green (61) found that 3 of 18 patients had fewer epileptiform discharges after 6 months of phenytoin treatment than did
control subjects, but this observation did not correlate with seizure frequency.
Phenytoin withdrawal has been associated with false localization of ictal onset (62).
Data regarding the effect of carbamazepine on interictal EEG discharges are conflicting. Most studies have found that carbamazepine either increases the amount of interictal abnormalities or has no effect. Studies by Jongmans (63), Wilkus et al. (22), and Sachdeo and Chokroverty (64) found that carbamazepine increased interictal epileptiform activity, but this was unrelated to seizure control. Wilkus et al. (22) used a randomized, crossover design to compare phenytoin and carbamazepine in 45 patients, most of whom had complex partial seizures. Focal epileptiform discharges increased significantly while patients were taking carbamazepine, but this did not correlate with seizure control. Pryse-Phillips and Jeavons (19) did not find any change in the amount of focal epileptiform activity when all patients were considered together, but 3 of 22 patients had increased epileptiform discharges while on carbamazepine; baseline rates in these patients were reestablished when the drug was stopped. In a double-blind, placebo-controlled trial involving 40 patients with psychomotor epilepsy, 50% of the patients had less, and 38% had more, interictal epileptiform activity while on carbamazepine. These findings, like those of others, did not correlate with seizure control (20). Similarly, Monaco et al. (65) found no consistent effect of carbamazepine on the EEG and no relation between any EEG change and seizure control. Finally, Martins da Silva et al. (58) reported a negative correlation between serum carbamazepine level and interictal epileptiform activity, but their study included only two patients on carbamazepine monotherapy.
The effect of valproate on epileptiform activity seems to depend on the type of epilepsy. Valproate clearly reduces generalized spike-wave discharges but probably has little or no effect on focal epileptiform activity.
Valproate suppresses 3-Hz spike-wave activity, and this correlates with control of absence seizures (25,39,66), including attacks present only during hyperventilation (24). Bruni et al. (24) studied patients with absence seizures before and after 1 year of valproate therapy. Fifty-seven percent of patients had the number of spike-wave paroxysms reduced by more than 75%; in one-third of these, spike-wave activity disappeared completely. Valproate was equally effective whether spike-wave paroxysms were of short or long duration. Thus, valproate suppressed over 75% of spike-wave discharges lasting longer than 3 seconds in 62% of patients, and 9 of 21 of these patients had the spike-wave activity eliminated completely. Both Villarreal et al. (25) and Bruni et al. (24) looked at clinical and EEG effects in the same patients at 10 weeks and 1 year after starting valproate treatment. Optimal seizure control was achieved as soon as therapeutic drug levels were achieved, but reductions in the amount of generalized spike-wave activity continued to occur up to 1 year later. Villarreal et al. (25) studied 25 patients with absence seizures treated with valproate. Most of these patients had other seizure types as well and remained on AED polytherapy during the study. Seventy-nine percent of the patients had reduced numbers of spike-wave discharges after introduction of valproate; 45% had spike-wave activity reduced by more than 75%. Reduced numbers of paroxysms lasting longer than 3 seconds correlated with control of absence attacks.
Valproate also suppresses photoparoxysmal responses. In one study, valproate reduced the number of patients showing photoparoxysmal responses by 31% (24). Harding et al. (67) studied 50 patients with photosensitive epilepsy before and during treatment with valproate. Valproate abolished photic-induced epileptiform activity in 27 patients (54%), and 12 patients (24%) were improved as defined by a greater than 75% reduction in the number of flash rates to which the patient was sensitive (the “sensitivity range”). Attenuation of the photoparoxysmal response may persist after an acute valproate dose has been cleared from the blood (68) and up to 3 months after discontinuing chronic therapy (67).
In contrast to the unambiguous effect of valproate on generalized spike-wave activity, studies of focal spike activity have not revealed any consistent changes with treatment. Thus, there are reports of both reduced (23) or unchanged (39) focal epileptiform activity after treatment with valproate.
EEG effects of ethosuximide have been less well studied than those of other common AEDs. Reports agree, however, that ethosuximide reduces generalized spike-wave discharges in approximately 50% of treated patients. Sato et al. (66) observed that ethosuximide abolished spike-wave discharges in the 12-hour telemetered EEGs of 6 of 11 patients (55%). Ethosuximide was slightly less effective than valproate in reducing the number of spike-wave discharges lasting 3 seconds or less. Ethosuximide also decreases 3-Hz spike-wave discharges induced by intermittent photic stimulation (67).
Of the newer AEDs, most is known about lamotrigine. Single-dose studies have shown a decrease in both interictal
spike activity (69,70) and photoconvulsive responses (69). In patients with refractory absence seizures, lamotrigine can suppress generalized spike-wave discharges (71). In a study of patients with localization related and generalized epilepsy that compared computerized EEG data before and 4 months after addition of lamotrigine, interictal epileptiform discharges were decreased in both frequency (42,72) and duration (42). Rarely, however, lamotrigine has been reported to activate new generalized epileptiform discharges associated with clinical seizures (73).
Felbamate reduces the number of sharp-slow-wave complexes in patients with Lennox-Gastaut syndrome (74). Vigabatrin decreases the frequency of interictal discharges in patients with intractable complex partial seizures, but this is unrelated to its efficacy in controlling seizures (75). In patients participating in a double-blind, add-on study of tiagabine, there was no effect on interictal epileptiform activity (29).
A few of the newer drugs can induce or increase myoclonus. For example, vigabatrin can induce myoclonus and generalized polyspike-and-wave complexes (76). In a double-blind study using a placebo, tiagabine was associated with the disappearance of epileptiform discharges in a few patients (29). Tiagabine also rarely induces new seizure types and can activate new epileptiform discharges (29). Gabapentin does not affect interictal epileptiform activity (72).
EFFECTS OF ANTIEPILEPTIC DRUGS ON SLEEP-WAKE CYCLES
Excessive daytime somnolence and sleep disturbances are common complaints among patients with epilepsy (77). The reasons for this often are unclear because it frequently is difficult to separate the effects of the disorder from those of its treatment. Available data suggest that AEDs affect sleep structure, but it is not now possible to draw a consistent picture of these effects. In some patients, AEDs improve sleep by decreasing the number of microarousals that result from seizure activity (78). Two studies have examined the effects of carbamazepine on normal control subjects (79,80) and found that sleep efficiency was increased and rapid eye movement (REM) density was decreased. A rigorous study of administration of carbamazepine to normal subjects and patients with newly diagnosed epilepsy suggested that reduction in REM by carbamazepine was significant only in patients with epilepsy (72). Drake et al. (81) used ambulatory EEG recordings made in patients' homes to study sleep patterns in 17 patients with focal or generalized epilepsy. Phenytoin, carbamazepine, and valproate each were taken by 5 patients; the other 2 patients took clonazepam. Phenytoin and carbamazepine were associated with decreased total sleep time and significant increases in sleep latency, number of arousals, and periods of wakefulness after sleep onset compared with valproate and clonazepam. Phenytoin increased the amount of stage 1 sleep and decreased stage 2 sleep. Patients taking carbamazepine had significantly less REM sleep than did those taking other drugs. There were no differences among the drugs in the amount of stage 3 or 4 sleep. Interpretation of these data is limited by the heterogeneous patient population and absence of a control or comparison group. Thus, it is not clear whether the observed effects were due solely to AEDs or whether different types of epilepsy affect sleep structure in different ways.
In a somewhat more rigorous study, Wolf et al. (82) used a single-blind protocol to investigate the effects of phenobarbital or phenytoin on sleep in 40 patients with epilepsy. Observations were made before and after therapeutic levels of each drug were achieved. Most of the patients were newly diagnosed and previously untreated; all were on monotherapy during the study. When taking phenobarbital, patients showed significant decreases in sleep latency. More time was spent in stage 2 and less in REM sleep, and interspersed time awake was decreased. Patients on phenytoin fell asleep more rapidly and spent relatively more time in stage 3 or 4 non-REM sleep than in stages 1 or 2. There were no differences between the drugs in the number of awakenings or total amount of REM sleep. Patients with generalized epilepsy tended to have shorter early REM cycles than did patients with focal epilepsy.
A similar study was performed comparing ethosuximide and valproate using a single-blind crossover design (83). Ethosuximide increased stage 1 sleep and decreased the amount of stage 3 or 4 sleep. REM sleep was increased in the early cycles but decreased in the later ones. Valproate's effects were limited to increasing stage 1 sleep and prolonging the first REM phase.
In other reports, phenobarbital has decreased nocturnal arousals, increased stage 2 sleep, and decreased REM periods (82,84,85). Phenytoin has been found to increase the number of arousals and periods of wakefulness (81,82), and also to have no appreciable effect on sleep (86). Information regarding carbamazepine is equally inconsistent. Thus, the drug has been associated with increased arousals, increased episodes of wakefulness, and decreased REM (81), as well as no detectable effects (86). Phenytoin, valproate, and ethosuximide may all increase stage 1 sleep (81,83).
Of the newer AEDs, gabapentin has received the most attention. In patients with epilepsy, gabapentin increased REM and reduced awakenings (87), although it is not clear if this related to improved seizure control. Rao et al. (88) found that gabapentin increased slow-wave sleep in normal control subjects. An add-on study of gabapentin and lamotrigine in patients with epilepsy showed that both drugs increased REM and sleep stability (72), although it is possible that these changes were due to decreased seizure activity (89).
EFFECTS OF ANTIEPILEPTIC DRUGS ON EVOKED POTENTIALS
Many investigators have reported changes in visual evoked potentials (VEPs) in patients with epilepsy using trains of binocular flash stimuli (90) or pattern-reversal stimuli (36,91), but it is difficult to know if observed changes are due to the epileptic disorder or to AEDs. In general, most studies have failed to demonstrate any effect on VEPs by valproate (92) or other antiepileptic agents (93). Martinovic et al. (94) found no differences in pattern-reversal VEPs from untreated patients and patients who were well controlled on valproate or carbamazepine compared with normal control subjects. Treated patients with poor seizure control had prolonged P2 (P100) latencies and N1P2 amplitudes that were not significant. Valproate does, however, affect VEPs in photosensitive patients (90,92,95). Faught and Lee (95) found that the latency of the P2 (P100) component of pattern-reversal VEPs was shorter in patients with photoparoxysmal responses and epilepsy than in control subjects. Latencies increased in patients successfully treated with valproate.
Phenytoin and carbamazepine, but not phenobarbital, valproate, or primidone, prolong interpeak latencies of somatosensory evoked potentials (SSEPs) and brainstem auditory evoked potentials (BAEPs). The magnitude of this effect is related to the plasma drug concentration (36,92, 96, 97, 98, 99, 100, 101). Zonisamide does not affect BAEPs or SSEPs at therapeutic dosages (102).
Vigabatrin has been causally linked to both clinical and asymptomatic constriction of the visual fields and other visual abnormalities, especially sensitivity (103, 104, 105, 106). Most studies have not found an effect of vigabatrin on VEPs in unselected patients (107) or after brief exposure (108), although some have demonstrated abnormalities with long-term exposure. Krauss et al. (106) studied four patients treated with vigabatrin for 2 to 40 months who experienced symptoms of visual field constriction or blurring. All had evidence of retinal dysfunction on electroretinography (ERG). Two had normal VEPs, one had prolonged VEP latency, and one had normal latencies but decreased P1 amplitude. In another study involving 39 asymptomatic patients who had taken vigabatrin for an average of 52 months (range, 28 to 78 months), 7 (18%) had bilaterally delayed VEPs; 5 of these also had reduced amplitudes (109). The abnormalities did not correlate with severity of clinical symptoms or changes in ERG. In a large study of 201 patients, VEPs were obtained before treatment and again after vigabatrin exposure for up to 24 months (110). There were no consistent changes for the entire group. Although five patients who remained on vigabatrin had prolonged P100 latencies at the end of the study, a similar number had abnormal VEPs before starting the drug. BAEPs and SSEPs do not seem to be affected by vigabatrin (108,110,111). Although vigabatrin can lead to a striking, bilaterally concentric contraction of the visual fields in some patients, the drug does not produce a consistent effect on VEPs. Indeed, VEPs usually are normal even in the presence of a significant visual field defect. In contrast, ERG abnormalities show a high degree of correlation with visual field loss (104,106). Two conclusions emerge: First, the electrophysiologic findings indicate retinal dysfunction, not abnormalities of central white matter, as the origin of the field deficits. Second, VEPs are not a useful screening tool for detecting asymptomatic visual field loss. ERG can detect and quantify retinal dysfunction, and it therefore may be useful in following patients. Both ERG and visual field abnormalities persist, even after the drug is discontinued.
ROLE OF THE ELECTROENCEPHALOGRAM IN TREATMENT DECISIONS
EEG findings can assist clinicians in determining that a historical paroxysmal event was a seizure. Epileptiform discharges (EEG spikes or sharp waves) are highly correlated with seizure susceptibility (112,113). Unfortunately, this specificity is not matched by a similar degree of sensitivity. In patients with established epilepsy, a single EEG demonstrates specific epileptiform abnormalities in 50% to 59% of cases. Repeated examinations increase the yield of positive findings (112,114), as do sleep and sleep deprivation (77,113). AEDs other than valproate usually do not change these percentages (112).
EEG findings also are helpful in classifying seizures as focal or generalized when clinical information is ambiguous, and in establishing an epilepsy syndrome diagnosis. Such information is important in making rational treatment decisions.
Initiation of Treatment
Although an EEG that shows epileptiform activity often is helpful in diagnostic formulations, treatment decisions require that EEG abnormalities also be characterized in terms of their epileptogenic significance, especially their predictive value in calculating risk of further seizures.
Most studies of first unprovoked seizures in adults have reported that the recurrence risk is increased if the EEG is abnormal. van Donselaar et al. (115) performed a prospective study of 165 patients who had had an unprovoked first seizure. All patients had a routine EEG performed after sleep deprivation. If no epileptiform activity was seen, a second EEG was obtained. Seizures recurred within 2 years in 83% of patients whose EEGs showed epileptiform discharges, but in only 12% of patients whose EEGs were normal (Figure 2.5). In patients whose EEGs showed nonepileptiform abnormalities, seizures
recurred in 41%. None of the patients were treated with AEDs before the second seizure.
FIGURE 2.5. Cumulative seizure recurrence rates based on electroencephalographic findings in 157 patients with idiopathic first seizures. (From van Donselaar CA, Schimsheimer R, Geerts AT, et al. Value of the electroencephalogram in adult patients with untreated idiopathic first seizure. Arch Neurol 1992;49: 231-237, with permission.)
Other studies have shown that EEG epileptiform abnormalities increase the risk of recurrence approximately twofold in patients who have had a single unprovoked seizure compared with similar patients with normal EEGs (116,117). A dissenting conclusion was reported by Hopkins et al. (118), who found no relation between EEG findings and risk of recurrence in 408 adult patients. Although their data suggested that fewer patients with abnormal EEGs remained seizure free, especially those with focal epileptiform abnormalities, analysis of recurrence risk at specific time points (3, 12, and 24 months) did not find these differences to be significant.
Children also have a higher risk of seizure recurrence if the EEG is abnormal at the time of a first unprovoked seizure. Camfield et al. (119) studied 168 children with an initial afebrile seizure. Seizures recurred in 68% of children with focal epileptiform discharges but in only 37% of children whose EEGs were normal. Recurrence rates were 45% to 64% in children with other kinds of EEG abnormalities. Other studies have found that any type of EEG abnormality increases the risk of recurrence in children after a first afebrile seizure (116,120).
Discontinuing Antiepileptic Drugs
EEG findings can assist in making decisions about discontinuing AEDs in treated patients whose seizures are in remission. The relation of EEG findings to risk of seizure relapse after drug withdrawal has been studied in both children (121, 122, 123, 124) and adults (125, 126, 127, 128). In a retrospective study involving 62 adult patients, epileptiform discharges at the time of drug withdrawal were 50% more common in patients who relapsed, although the finding did not reach statistical significance (128). In a prospective study of 1021 patients, all but 28 had at least one EEG before drug withdrawal. Univariate analysis showed that only patients who had had generalized tonic-clonic seizures and EEGs showing generalized spike-wave abnormalities had a higher risk of seizure relapse; EEG findings were not predictive for other groups (127). In another well designed, prospective investigation, Callaghan et al. (126) studied 92 patients with generalized or complex partial seizures that had been controlled by AEDs for at least 2 years. EEGs were obtained before treatment was started and again before drugs were withdrawn. Patients whose EEGs were normal or improved with treatment had a 94% to 99% reduction in risk of relapse compared with patients whose EEGs were abnormal before treatment and unchanged before withdrawal (Table 2.1). Despite differences among available studies, we conclude that persistent EEG abnormalities increase the risk of seizure relapse after drug withdrawal in adults whose seizures have been in remission for 2 years or more.
There is considerably more agreement regarding EEG findings and seizure relapse in children. Virtually all studies have shown that EEG abnormalities are a major risk factor for seizure relapse after AEDs are discontinued. This association holds regardless of whether the EEG was recorded during the year before AED withdrawal (122), immediately before withdrawal (121,123,124), or after withdrawal (123). The most convincing study is that of Shinnar et al. (124), who prospectively investigated 88 children with afebrile seizures who had been seizure free on AEDs for at least 2 years. EEGs were obtained before drug withdrawal; earlier EEGs were available for comparison in 82 patients. EEG characteristics were strongly predictive of outcome after drug withdrawal: Seizure recurrence was substantially lower in children who had EEGs that were normal or improved at the time drugs were discontinued. Specific EEG features (slow-frequency activity,
spikes) were more informative than classification only as normal or abnormal (Figure 2.6).
TABLE 2.1. ELECTROENCEPHALOGRAPHIC FINDINGS AND RELAPSE RATES
FIGURE 2.6. Effect of slowing and spikes on the probability of remaining free of seizures over time. (From Shinnar S, Vining EPG, Mellits ED, et al. Discontinuing antiepileptic medications in children with epilepsy after two years without seizures: a prospective study. N Engl J Med 1985;313:976-980, with permission.)
Antiepileptic drugs have significant effects on brain physiology as measured by evoked potentials, waking-sleep patterns, and, especially, the EEG. EEG data are helpful in deciding whether to initiate or discontinue treatment. AEDs, especially at toxic levels, often affect EEG background activity, and this may be an important interpretive consideration in some patients. In some circumstances, AEDs may help clarify EEG findings, as in distinguishing primary from secondary bilateral synchrony, or aid in localization of the epileptogenic zone. Whether computer-assisted methods of spike analysis may increase the utility of interictal EEG findings as a gauge of AED efficacy remains to be proved. Finally, AEDs affect waking-sleep cycles and sleep structure, but these associations need further characterization to be clinically useful.