Michael Koutroumanidis and Carmen Martin-Miguel contributed equally to this paper.
Interictal Temporal Delta Activity in Temporal Lobe Epilepsy: Correlations with Pathology and Outcome
Article first published online: 27 OCT 2004
Volume 45, Issue 11, pages 1351–1367, November 2004
How to Cite
Koutroumanidis, M., Martin-Miguel, C., Hennessy, M. J., Akanuma, N., Valentin, A., Alarcón, G., Jarosz, J. M. and Polkey, C. E. (2004), Interictal Temporal Delta Activity in Temporal Lobe Epilepsy: Correlations with Pathology and Outcome. Epilepsia, 45: 1351–1367. doi: 10.1111/j.0013-9580.2004.61203.x
- Issue published online: 27 OCT 2004
- Article first published online: 27 OCT 2004
- Accepted June 20, 2004.
- Interictal EEG;
- Regional slow activity;
- Temporal lobe epilepsy;
- Prognostic value
Summary: Purpose: To determine the characteristics and the clinical significance of focal slow activity and its association with focal epileptogenesis in patients with temporal lobe epilepsy (TLE).
Methods: We analyzed the interictal EEGs of 141 patients who had temporal lobe resections for intractable focal seizures and correlated the findings with pathologic changes and outcome. The pathologic changes were categorized into medial temporal sclerosis, tumors, and nonspecific changes.
Results: Lateralized slow activity was found in 66% of the patients, and it was mainly temporal, of delta frequency and irregular morphology. None of its characteristics, including quantity and reactivity to eye opening, was substrate specific. It was highly concordant with temporal spiking (60%), without any difference across the three groups, but provided additional information in 19 (15%) patients who had no lateralizing spikes. The effect of sleep also was similar in all three groups and included transition of slow waves into spikes. Lateralized slow activity to the side of the operation was significantly associated with favorable outcome only in the group with nonspecific pathology (p = 0.008), regardless of the presence, laterality, or topography of spikes.
Conclusions: Our findings suggest that in patients with TLE whose brain magnetic resonance imaging (MRI) is either normal or suggestive of medial temporal sclerosis, interictal temporal slow activity has a lateralizing value similar to that of temporal spiking. Its association with a favorable outcome in patients with nonspecific pathology also suggests that candidates with lateralizing temporal delta and normal MRI should not be barred from further preoperative assessment.
Regional slow activity was first recognized by Grey Walter (1) in his seminal work on patients with cerebral hemisphere tumours and was termed delta because of its association with disease, degeneration and death. Regional delta activity has been subsequently assigned as the most common EEG abnormality associated with underlying mass lesions (2,3) and was thought to reflect anatomical deafferentation of the overlying cortex (4,5). In the meantime, other clinical/EEG studies showed that focal slow activity also could occur in other disease states, either persistently, as in cerebrovascular accidents and related edema (6), or transiently, after a complex migraine attack (7).
In focal epilepsy, regional slow activity was thought to reflect either postictal changes or an underlying structural lesion until the early era of the modern structural imaging, when it became clear that such activity may occur in the absence of abnormalities in brain computed tomography (CT) scans (8,9). Kaibara and Blume (10) explored the time scale of the regional postictal electrographic changes, and consequently the concept of the interictal regional slow activity in chronic focal epilepsy was refined. Currently, interictal regional slowing and attenuation of normal fast biologic activities constitute the electrographic evidence of the “functional deficit zone,” which is defined as the cortical area of nonepileptic dysfunction (11). Requiring no barbiturate or benzodiazepine (BZD) administration, interictal regional slow activity is easier to study, and indeed, a number of studies over the last 15-year period have explored its usefulness in the lateralization and localization of the epileptic focus (12–19). However, no agreement has been reached so far on the incidence, usefulness, and limitations of the interictal regional slow activity in focal epilepsy, with the different methods used (criteria for lateralization, visual vs. spectral analysis, consideration of rhythmic or arrhythmic patterns, and different patient-selection criteria) being—at least partly—responsible for the observed discrepancies. For example, the incidence of the interictal regional slow activity has varied widely between studies from 100% (12) to just 25% (20) by using visual analysis, and from 50% (20) to 90% (15) by using power spectra. Some investigators have considered only rhythmic delta (14), and others, only continuous patterns (21). Whereas most authors agree that interictal regional slow activity can reliably lateralize the epileptogenic side, others have reported occasional false lateralization (15,16).
In contrast to the “epileptiform” spikes, the biologic significance of the interictal slow activity is still uncertain. Electroencephalographically, the concept of the “functional deficit zone” is poorly understood, and everyday practice in EEG reading shows that regional slow activity and loss of fast rhythms can occur independently, suggesting different pathophysiologic mechanisms. Anatomic cortical deafferentiation (5) may well be the principal underlying mechanism, at least in patients with space-occupying lesions, but it is uncertain whether and to what extent it also can explain the interictal appearance of regional slow activity in patients with atrophic pathology such as medial temporal sclerosis (MTS), with which it is closely associated (15,17,18). Some investigators have hypothesized that in nonlesional focal epilepsy, interictal regional slow activity may represent a genuinely epileptiform pattern (17,18), but this hypothesis should be tested in patients without demonstrable pathology, as some degree of anatomic deafferentiation of the overlying cortex may be expected in MTS, at least theoretically, because of variable hippocampal neuronal loss.
The relevance of the morphology of the interictal regional slow activity also is uncertain. Although some investigators have recognized rhythmic and arrhythmic subtypes, and have assigned different clinical significance to each type (14,19,21), others seem to have ignored such a distinction (15–18,22). Finally, whereas a number of studies have addressed the predictive value of interictal spikes in terms of seizure outcome after resective epilepsy surgery [and a consensus has been reached that prognosis is generally favorable when interictal spikes are restricted to the surgery focus (23–28)], only one study has considered regional slowing, which found that lateralized temporal delta activity to the later-resected temporal lobe had no significant bearing on outcome (29).
In this study, we examined the clinical, EEG, imaging, and histopathologic data of 141 patients with temporal lobe resections for intractable focal seizures, seeking to determine (a) the incidence, lateralizing, and localizing value of interictal regional slow activity and its lateral concordance with the regional spiking; (b) whether it is specific for a given pathologic substrate (MTS and tumors), or can overlie other conditions, including nonspecific (NS) changes; (c) its inherent variability in terms of morphology, frequency, quantity, and reactivity, and whether this may differ according to the type of the underlying pathology; and (d) its predictive value regarding seizure outcome after resective surgery. We chose to set about by using the term interictal regional slow activity (IRSA), as this defines with sufficient clarity the fundamental characteristics of the pattern (timing in relation to seizure occurrence and restricted topography) and allows the study of its inherent variability and clinical relevance.
PATIENTS AND METHODS
All patients who underwent resective surgery in the temporal lobe between 1975 and 1997 at the Maudsley and King's College Hospitals, London, were reviewed, and those meeting all of the following inclusion and exclusion criteria were consecutively selected for the study. Inclusion criteria were (a) medically intractable focal seizures, compatible with a temporal lobe origin (30), and (b) follow-up for ≥2 years after resection. Exclusion criteria were (a) history of prior brain surgery or fixed neurologic deficit of central etiology at the time of presurgical evaluation (which could be a cause for regional slowing); (b) clinical or radiologic evidence of a rapidly progressive brain lesion, such as a malignant brain tumor (which might suggest another reason for operation apart from medical intractability of epileptic seizures); however, patients with histopathologically confirmed low-grade glial or dysembryoplastic neuroepithelial tumors (DNETs) were included; and (c) significant postoperative complications suggesting additional destruction of brain tissue (that might imply other reasons apart from incomplete removal of the epileptogenic area as a cause of postoperative seizures, which in turn would blur the association of EEG variables with outcome). Minor complications, such as quadratic visual field defects, third-nerve palsy, or transitory anomia (31) were accepted.
We required at least one awake and sleep, or two awake preoperative interictal EEG recordings for each patient. The recency and the type of the last seizure were noted from the contemporary technical report, and EEGs were considered interictal when they contained no electrographic seizure and were recorded at least half an hour after the last complex partial seizure (CPS), or 1 h after the last generalized tonic–clonic seizure (GTCS) (10,18). Patients with postictal recordings or traces lacking information on seizure recency were excluded. To eliminate false-positive findings related to the physiologic posterior slowing in children, adolescents, and some young adults (32,33) and the posterior slowing of young children during light sleep (34), we chose to use the following combined criteria instead of arbitrary cutoff age limits: (a) prominent temporooccipital or parietooccipital topography intermixed with or underlying the alpha activity with similar voltage and reactivity (i.e., blocking with eye opening and disappearance during drowsiness and light sleep), and (b) consistent interside asymmetry <50% (35). Children were considered separately. At the time of preoperative assessment, all patients were treated with an optimal regimen of antiepileptic drugs (AEDs), and none had any neurologic condition other than epilepsy, including complicated migraine, or any acute or chronic medical illness.
From a series of 350 consecutive patients, who met all criteria, we were able to retrieve truly interictal, good-quality, preoperative scalp EEGs in only 141 (81 male and 60 female) patients. In total, 212 traces were analyzed (1.5 recording per patient). In 116 patients, recordings containing variable sleep stages were obtained for a variable time by oral administration of 100 to 200 mg of quinalbarbitone in adults, or trimeprazine tartrate, 2 mg/kg, in children, whereas 12 others slept spontaneously. For the remaining 13 patients, at least two standard awake recordings were available. Mean recording time per patient was ∼70 min. All EEGs incorporated hyperventilation and photic stimulation and were recorded by using the Maudsley electrode-placement system (36). A 21-channel Nihon-Kohden electroencephalograph was used for most of the cases, but a few of the earliest recordings were taken on a 16-channel machine. Common average referential and longitudinal and transverse bipolar montages were used in all examinations and for most of the recording time. None of these EEGs was part of prolonged telemetry, and all patients were taking full AEDs at the time of the recordings.
All EEGs were visually analyzed by two electroencephalographers (M.K. and C.M.M.), who at this stage were blind to all other preoperative test results, side of operation, histopathology, and outcome, with any differences resolved by consensus. Spikes and IRSA showing a 4:1 (80%) or greater ratio of predominance on one side (in terms of frequency and time of appearance, respectively) were considered lateralized (13) and, for the subsequent analysis, they were grouped together with those showing strictly unilateral spiking and slowing, as “lateralized.” Two other categories were formed for both spiking and IRSA: “bilateral, nonlateralized” and “absent abnormality.” The topography of spikes and abnormal slow activity was classified according to the involved electrode positions as temporal (when restricted mainly within this lobe), temporal-plus (when mainly temporal but expanded to other lobes), extratemporal (when mainly frontal or parietal), hemispheral (when they occupied half of the brain), or generalized (bursts occupying the whole of the cerebrum), and the area of the maximum voltage was noted. For every recording, all different types of abnormal slow activity were considered (for example, coexistence of focal and bilateral synchronous bursts), and the effects of hyperventilation and sleep also were noticed when possible. In addition we studied the following parameters:
- 1Morphology. Slow activity was characterized as polymorphic or arrhythmic when slow waves were irregular in shape and had a variable duration without a stable predominant frequency, and as monomorphic or rhythmic when they were regular in shape with fairly constant duration and predominant frequency (37). The terms polymorphic and monomorphic shall be used as better describing the morphology of the pattern.
- 2Frequency (predominant theta or delta).
- 3Abundance, defined as time of appearance in respect to the total duration of the recording, was classified on a 3-point scale (<25%, 25–50%, and >50%).
- 4Reactivity. Because of the retrospective nature of the study, reactivity could be assessed only with respect to eye opening, either alone or in association with photic stimulation. It was classified as “full” when ongoing focal slowing was clearly and consistently abolished on eye opening and for as long as the eyes remained open; “partial,” in case of attenuation or when the effect was not sustained for the whole “eyes open” epoch; and “none,” when eye opening produced no clear change. Recordings with focal slowing of low abundance and, by implication, with little chance to occur during the relatively brief periods of eye opening, were not considered in this respect.
Clinical data and neuroimaging
The epilepsy and medical history, operation note, pathology, and outcome were assessed from contemporary and follow-up clinical records. Patients with conditions that could produce focal slowing, such as previous brain surgery, cerebrovascular accident sustained after the onset of their epilepsy, or history of complicated migraine, were excluded at this stage. Seizure frequency was evaluated on 3-point scale: monthly (one to three seizures per month), weekly (one to six seizures per week), or daily (more than six seizures per week). Preoperative and postoperative neuroimaging was performed in all patients. Sixty-seven had brain magnetic resonance imaging (MRI) on both occasions, whereas the others who were assessed before MRI became available had CT scans. Sixteen of the latter patients had also postoperative MRI scans as part of their reassessment due to seizure recurrence. In total, 83 (60%) patients had MRI at least once, whereas 37 patients had both MRI and CT scans. All CT reports were reviewed, but only 30% of the actual films were available for reevaluation. All MRI scans were reevaluated.
Surgery, histopathology, and outcome
All patients were operated on by the same neurosurgeon (C.E.P.); 132 had anterior temporal lobectomy (73 right and 59 left), and nine, amygdalohippocampectomy (three right, six left). In dominant lobectomies, the resection included 4.5 to 5 cm from the temporal tip, whereas in nondominant, it measured 5 to 5.5 cm from the tip. The amygdala and the anterior two thirds of the hippocampus were removed. Amygdalohippocampectomies were performed through a transsylvian route (38). In general, the decision for operation was reached when congruent data were available from the clinical history and the neurophysiologic, neuroimaging, and neuropsychological investigations. Because the present cohort were assessed and operated on over a long period (1975–1997), preoperative assessment was not uniform for each case. Nevertheless, all patients with nonlateralizing imaging or extracranial EEG underwent intracranial recordings. Regional slow activity in the interictal scalp EEG had been invariably regarded as “background disturbance” and had been assigned at the time of the preoperative investigation the same significance as regional attenuation of fast rhythms.
All temporal lobe specimens included the medial temporal structures (amygdala, hippocampus, and parahippocampal gyrus). The neuropathologic reports of the excised specimens were reviewed, and MTS was considered present if the original report had specified neuronal loss and gliosis in the hippocampus with the most severe changes in the Sommer sector (H1) and the end-folium (H3–5) and less severe changes in the dentate fascia and the resistant zone (H2) (39,40). Minor degrees of white matter neuronal ectopia, although always noted, were not considered to represent dual pathology. Quantitative techniques were not used. The original slides of patients with low-grade tumors were reviewed because of the relatively recent emergence of DNET as a diagnostic entity (41,42), and the completeness of the histologic resections was ascertained. Pathology was deemed nonspecific if no structural lesion was present and the classic appearances of MTS were absent (43). Minor degrees of gliosis confined to the end-folium also were considered nonspecific and likely to be related to the consequences of chronic seizure activity rather than a primary epileptogenic lesion (44). Pathologic findings were categorized as MTS, tumor, nonspecific (NS) pathology, and “other,” which included cortical dysplasia, arteriovenous malformation, and arachnoid cyst.
Outcome data were prospectively collected in regular postoperative follow-up clinical assessments and were available for all 350 patients of this series. The 141 patients of this study were divided into groups according to the most recent outcome state by using Engel's classification (23). For analysis, outcome was evaluated on a 2-point scale: favorable (classes 1 and 2) and poor outcome (classes 3 and 4), and seizure free (class 1 only) and non–seizure free (classes 2–4) (45) to avoid small sample sizes.
The median age at seizure onset was 8 years (range, 6 months to 39 years), the median disease duration was 13 years (range, 2–37 years), and the median age at EEG examination was 21 years (range, 3–56 years, with 14 patients younger than 13 years). All patients had CPSs, which at the period of the presurgical assessment occurred daily in 54, weekly in 51, and monthly in 20. Seizure frequency was unknown in 16 patients. Secondary GTCSs occurred in 65 of 136 patients in whom relevant information was possible to retrieve. Neurologic examination was normal in all.
MTS was identified in 53, NS changes in 32, and tumors in 46 patients. In the latter group, review of the original slides showed DNET (42) in 38 patients, low-grade ganglioglioma in four, hamartoma in two, intrinsic tumor in one, and a calcified mass in one patient. On microscopic examination, DNET involved the medial temporal areas in three patients and coexisted with MTS in other 10 patients. One patient with ganglioglioma also had MTS. Tumor resection was incomplete (i.e., abnormal tissue at the margin of the specimen) in 26 patients, complete in 15, and five specimens were fragmented.
Of the 10 remaining patients, five had focal cortical dysplasia (46), two had grey matter heterotopias, two had arteriovenous malformations, and one had an arachnoid cyst. The type of pathology was not related to the duration of epilepsy, seizure frequency, or occurrence of GTCSs.
CT scans were performed on 16 patients and showed unilateral temporal atrophy or dilatation of the lateral ventricle in seven (contralateral to the resection in three), diffuse atrophy in one patient, and were normal in eight. MRI was performed in 41 patients; it showed hippocampal formation atrophy, or increased signal on T2-weighted images, or both ipsilateral to the resection in 34 patients, and was normal in seven. No other abnormality was noted in any of the scans.
CT scans, performed in 43 patients, showed a lesion in 34 (hypodense in 12, calcified in 22), ipsilateral atrophy in three, and were normal in six patients. MRI was performed on nine patients and provided a detailed anatomic delineation of the tumors. Lesions were unilateral and solitary and were confined to the temporal lobe in all but one patient, in whom a calcified lesion (DNET) involved mainly the uncus but extended to the ipsilateral thalamus. Scans were otherwise unremarkable.
NS pathology group
CT scans, performed in 25 patients, showed dilatation of the lateral ventricle in two, ipsilateral superior temporal gyrus atrophy in one, and were normal in 22 patients. MRI was performed in eight patients and was normal in all.
Of the 16 patients who had MRI only postoperatively, histopathology had shown MTS in two, DNET in six, and NS changes in eight. With the exception of the two MTS patients who had contralateral hippocampal signal changes without atrophy, no apparent changes were seen in the hemisphere opposite the resected temporal lobe. On the side of the operation, incomplete resection was noted in three patients with DNET and occipital atrophy in one patient with NS pathology. No other abnormalities were observed.
Follow-up and outcome
The mean follow-up period was 53 months (median, 38; range, 24–169 months). Outcome was favorable (Engel's classes 1 and 2) in 94 (66.7%) patients; 40 (75.5%) with MTS, 34 (77.3%) with tumors, and 12 (37.5%) with NS changes. Detailed outcome data for all patients is presented in Table 1.
Lateralization and topography of interictal abnormalities
Slow activity IRSA was lateralized (including unilateral) in 93 (66%) patients, nonlateralized in 27 (19%), and absent in 21 (15%) patients. Table 2 shows the lateralization and topography of slow activity for each major diagnostic category (MTS, tumors, and NS). Lateralized IRSA was noted contralateral to the operation in five (3.5%) patients. Bilateral independent nonlateralized temporal slowing (Figs. 1 and 2) was noted in 21 patients (see Table 2), six of whom had unilateral tumors (patients 3 through 8 in Table 3). Twelve patients had brief bilateral bursts of rhythmic or irregular, high-voltage, and usually anteriorly predominant slow activity (Table 2).
|MTS (n = 53)||Tumors (n = 46)||NS (n = 32)|
|Lateralized||31 (58.5%)a||34 (74%)b||21 (65.5%)b|
|Nonlateralized||14 (26.5%)||8 (17%)||3 (9.5%)|
|Bil. synchronous bursts over all areas||5d||6e||1d|
|None||8 (15%)||4 (9%)||8 (25%)|
|Maximal amplitude (lateralized)|
|Ant. to midtemporal||24||19||16|
|Lateralized||32 (60%)a||28 (61%)a||19 (59%)c|
|Nonlateralized||7 (13.5%)||9 (19.5%)||6 (19%)|
|Bil. synchronous other than temporal||5f||11g||2|
|None||14 (26.5%)||9 (19.5%)||7 (22%)|
|Maximal amplitude (lateralized)|
|Ant. to midtemporal||25||18||11|
|Pts||Age (yr)/ duration||Seizures/ mo||Slow: side/ topography||Slow: time of appearance (%)||Spikes: side/ topography||Neuroimaging (CT/MRI)||Operation||Pathology/ resection||FU (mo)||Outcome|
|1||42/27||8||R/temp||50||R/temp||L temporal lesion (CT)||LTL||GG/incomplete||53||1B|
|2||22.5/22||1||R/temp||10||None||Calcified, nonenhancing lesion at L uncus, growing along medial temporal structures to ipsilateral thalamus (CT)||LTL||DNET/incomplete||49||1A|
|3||45/22||3||BI/temp||25||R>L/temp||R mes. temporal calcified lesion (CT)||RTL||DNET/complete||169||1D|
|4||6/5||12||BI/temp plus||25||R/temp||Normal (CT)||RTL||DNET/incomplete||12||1A|
|5||17/5||8||BI/temp plus||25||None||L mes. temporal calcified lesion (CT)||LTL||DNET/incomplete||154||3A|
|6||18/9||90||BI/temp plus||25||BI and BS/temp||Normal (CT)||LTL||DNET/incomplete||155||3A|
|7||18/3||10||BI/temp||50||BI/temp||L temporal calcified lesion (CT)||LTL||DNET/complete||35||1A|
|8||28/7||12||BI/temp plus||25||None||R temporal lesion (CT)||RTL||GG/complete||74||2A|
|9||18/5||Unknown||BI/temp||50||None||T2 hyperintense lesion in L inferior temporal gyrus (MRI)||LTL||DNET/fragmented||87||1C|
|10||12/5||16||R>L/temp||90||BI/temp||R mes. temporal lesion (CT)||RTL||DNET/incomplete||35||1A|
Spikes Focal spiking was lateralized (including unilateral) in 85 (60%) patients, nonlateralized in 25 (18%), and absent in 31 (22%) patients (Table 2). Lateralized focal spikes were noted contralateral to the operation in five (3.5%) patients.
Associations between IRSA and spikes In total, lateralized spikes or focal slow activity, or both, were observed in 106 (75.2%) of the 141 patients, and their association is shown in Table 4. Nineteen patients had lateralized IRSA to the side of subsequent operation but either nonlateralizing spikes or no clear spiking at all. Eight of these had DNET (six seizure free), four MTS (three seizure free), six NS changes (four seizure free), and one arteriovenous malformation (seizure free).
|Temporal spiking||Temporal slowing|
|Correctly lateralized (n = 89)||Nonlateralized/absent (n = 47)||Falsely lateralized (n = 5)|
|Correctly lateralized (n = 81)||70||11||0|
|Nonlateralized/absent (n = 55)||19||35||1|
|Falsely lateralized (n = 5)||0||1||4|
Six (4.3%) patients had lateralized focal slowing and spiking contralateral to the operation. Both spikes and temporal slowing were discordant in the patient with MTS and in two of the three patients with NS pathology, with the third one showing no slowing (Table 2). One of the two patients with tumor and contralateral focal slowing also had false lateralizing temporal spikes (Fig. 3); the second showed no epileptiform activity in two recordings (patients 1 and 2 in Tables 2 and 3).
Both IRSA and regional spiking were well restricted within, or primarily involved, the temporal electrodes in the vast majority of patients with lateralized findings (Table 2).
No difference regarding lateralization and topography was noted between focal spikes and slow activity either in each group or in the overall patient population. The presence of focal slowing or spiking for all patients as well as separately for each group was not related to age at onset, duration of epilepsy, age at EEG examination, seizure frequency, or GTCS occurrence.
Amplitude, morphology, and frequency of lateralized IRSA and spiking
The maximal voltage of lateralized slow activity often switched topography within the longitudinal temporal axis and even some neighboring extratemporal (usually the superior frontal or sylvian) areas in the same EEG, but it was recorded predominantly over the anterior temporal electrodes in the majority of the patients (70%). The actual voltage was ≤80 μV in most of the patients (24 with MTS, 20 with tumors, and 17 with NS), and higher (≤150 μV) in the others. Lateralized monomorphic slow activity was noted in 32 (37%) patients, mainly intermixed or alternating with bursts or runs of polymorphic slow activity of the same topography (27 patients, 31%), and only rarely seen as the predominant interictal pattern (remaining five patients, 6%). In contrast, the polymorphic variety was the predominant pattern of IRSA in 54 (63%) patients. The frequency of the IRSA was mainly within the delta range, with only a few recordings showing predominant or significantly intermixed theta rhythms. No significant difference was noted between the three diagnostic categories regarding amplitude, morphology, and frequency of lateralized focal slow activity (Table 2).
The maximal voltage of the lateralized spikes showed a topography similar to that of the slow activity and was not different among the three groups (Table 2).
Abundance and reactivity of lateralized IRSA
The quantity of correctly lateralized slow activity was significantly greater in patients with tumors (Kruskal–Wallis test, χ2= 6.077; p = 0.048). Conversely, correctly lateralized IRSA was >50% in seven (33.4%) of 21 patients with NS changes and in a similar proportion of patients with MTS (11 of 31, 35.5%).
Reactivity to eye opening was possible to assess in 54 patients with lateralized IRSA. Patients with tumors showed less-reactive temporal delta, although four clearly showed either partial (two patients; Fig. 4) or full reactivity. Of note is the observation that IRSA was clearly not reactive in eight of 12 patients with NS changes (Fig. 5). Finally, some degree of reactivity was noted in ∼50% of patients with MTS.
Effects of drowsiness and light sleep on lateralized IRSA and spiking
Sixty-nine patients with lateralized IRSA (28 with MTS, 24 with tumors, and 17 with NS pathology) had recordings during wake state, drowsiness, and light sleep of sufficient duration for reliable interpretation. During drowsiness and light sleep, IRSA increased in quantity or appeared for the first time in 34 (51%) patients (18 with MTS, 13 with tumors, three with NS), and attenuated or disappeared in nine (13%) patients (four with MTS, three with tumors, two with NS). In three of these nine patients (one from each group) who had no spikes during wakefulness, IRSA gave way to spikes of the same topography during sleep. Other effects of drowsiness and light sleep on IRSA included increase in voltage and field of distribution (the latter became wider, from temporal to anterior quadrant and even hemispheral) in two patients, decrease in frequency (from predominantly theta to delta) in all four patients with theta activity (Table 2), and morphologic changes in four patients (from polymorphic into monomorphic in two, and vice versa in the two others).
Lateralized spiking increased or appeared for the first time during drowsiness and light sleep in a similar proportion of patients (29 of 53, 44%), but did not decrease in any.
Relation of IRSA and interictal regional spiking with outcome
IRSA, when lateralized to the side of the operation, was significantly associated with favourable outcome (Engel's classes 1 and 2; Table 5a) and seizure freedom (Engel's class 1; Table 5b) only in the group with NS pathology (χ2= 6.296; p = 0.008; sensitivity, 91.7%; 95% CI, 65.9–99.6; specificity, 60.0%; 95% CI, 44.6–64.7; positive predictive value, 57.9%; 95% CI, 41.6–62.9; negative predictive value, 92.3; 95% CI, 68.6–99.6 with respect to the favourable/poor outcome measure, and χ2= 3.960; p = 0.024; sensitivity, 90.0%; 95% CI, 59.8–99.5; specificity, 54.5%; 95% CI, 40.8–58.8; positive predictive value, 47.4%; 95% CI, 31.5–52.3; negative predictive value, 92.3; 95% CI, 69.1–99.6 with respect to seizure-free/non–seizure-free outcome measure). In this group, unilateral or lateralized focal spiking on its own right was not significantly associated with favorable or seizure-free outcome measures. No significant relation between lateralized spikes or focal slowing and outcome was found in the groups of patients with MTS and tumors, or in the overall patients.
|Temporal slow||Temporal spikes|
|Correctly lateralized||Elsea||Statistics [χ2 (p)]b||Correctly lateralized||Elsea||Statistics [χ2 (p)]c|
|All patients(n = 141)|
|MTS (n = 53)|
|Tumours (n = 46)|
|NS (n = 32)|
|Temporal slow||Temporal spikes|
|Correctly lateralized||Else||Statistics [χ2 (p)]b||Correctly lateralized||Else||Statistics [χ2 (p)]c|
|All patients (n = 141)|
|MTS (n = 53)|
|Tumours (n = 46)|
|NS (n = 32)|
|Non–seizure free||10||12||3.960 (0.024)||10||12||0.145|
Findings in children
Thirteen children aged from three to 13 years (median, 9 years; SD, 3 years) were included in the study. Histopathology showed tumor in 12 (10 had DNET, one had ganglioglioma, and one had intrinsic tumor) and MTS in one. Outcome was favorable in 10. Twelve showed lateralized IRSA, which was restricted to the temporal areas in four, but extended to the frontal and central areas in six, to the whole hemisphere in one, and to the posterior quadrant in another one child; one had bitemporal, independent, nonlateralizing slowing (patient 4 in Table 3). IRSA occupied more than half of the recording time in 10 children (in five of them, it was almost continuous), was mainly polymorphic (but strictly monomorphic in one); its frequency was invariably within the delta range (<2 Hz in three children), and its voltage was generally higher than that in the adults (>100 μV in 10 children). It was nonreactive in six and partially reactive in three, including the child with MTS; in the remaining four children, reactivity to eye opening was not possible to assess. Eight children had ipsilateral spikes, which were of temporal distribution in six and hemispheral in two; three had diffuse bilateral spiking; one had bitemporal independent spikes; and one had no spikes.
We report the interictal EEG findings of 141 patients who underwent temporal lobe resections for medically intractable CPSs, focusing primarily on the IRSA and, in particular, on the clinical significance and the intrinsic relation of this pattern to focal epileptogenesis. Particular strengths of this study are the relatively large number of awake and sleep recordings despite the rigorous exclusion criteria applied, and that EEG findings were correlated with pathology rather than with imaging. Furthermore, this is the second study to the best of our knowledge that assessed the prognostic value of the IRSA, and the first to have done so in different histopathologic substrates.
We found that IRSA is at least as useful as the interictal temporal spiking in lateralizing the epileptogenic area and is a good predictor of postoperative seizure relief in patients with TLE associated with mild NS pathology. A high lateral concordance of IRSA was found with spikes in all groups, and both patterns showed remarkably alike behavior during drowsiness and sleep. Incidence, lateralization, morphology, and reactivity of IRSA were not pathology dependent. Apparently paradoxic findings include its occurrence contralateral to temporal tumors and tumor-like characteristics (such as appearance in large quantities and lack of reactivity on eye opening) in patients without demonstrable specific histopathologic changes. Before discussing our findings in details, we must raise the following methodologic issues.
Meaningful interpretation of our findings, particularly when IRSA was contralateral to the resected temporal lobe or bilateral independent (five and 21 patients, respectively, Table 2) would require exclusion of structural lesions contralateral to the operation. All patients had preoperative and postoperative neuroimaging, but a number of them had only CT scans, as they were operated on before the MRI era. CT scanning failed to detect DNET in six of our 43 patients with tumours, and similar has been the experience of others (47); therefore it can introduce “false negative” errors. MRI effectively excluded contralateral structural lesion in 83 (60%) patients, of whom 10 had contralateral (one with NS pathology) or substantial bilateral independent (six with MTS, one with DNET, and two with NS pathology) temporal slowing. Of the remaining 16 patients with contralateral (four patients) or substantial bilateral independent (12 patients) temporal slowing who had only CT, 11 [seven with DNET (Table 2), three with MTS, and one with NS pathology] had good outcome (thus ruling out any contralateral epileptogenic structural lesion undetected by CT), leaving only five patients (two with MTS, two with DNET, and one with NS pathology) with poor outcome and contralateral slowing unaccounted for on imaging grounds. The possibility of all these five patients harboring epileptogenic structural lesions missed by CT is remote, and at any rate, it would not alter the principal findings of this study.
To establish a reliable relation between electrophysiologic values and outcome, we chose to exclude patients with <2 years' follow-up, as the majority (≤86%) of recurrences take place within this period (48), and little change has been shown to occur in the patient's seizure status thereafter (25,49,50). For accurate electropathologic correlations, the original slides of patients with low-grade tumours were reviewed, and a number of them were reclassified as DNET (42), and in the vast majority of patients of the “tumor” group, the completeness of histologic resection was ascertained.
Incidence and lateralizing value of interictal regional slow activity and spiking in TLE
The similarities between IRSA and regional spiking in terms of incidence and lateralizing value, regardless of the type of the underlying pathology (Tables 2 and 4), are impressive, and in accordance with previous experience (16,17). In total, lateralized IRSA was present in 93 (66%) patients, and in spite of the high lateral concordance with spikes, it provided additional useful information in 19 (13.5%) patients with absent or poorly lateralizing focal spikes. The presence of lateralized temporal slowing was particularly important for the six patients with NS pathology as the only solid lateralizing parameter; brain MRI was, or presumed in those with CT only, normal. Four (two thirds) became seizure free after ipsilateral anterior temporal lobectomy.
The incidence of false lateralization (for the side of operation) also was remarkably similar between IRSA and spiking in our material (3.5%); interestingly, spikes also were discordant in four of the five patients with falsely lateralizing IRSA, including one patient with a left temporal ganglioglioma (Table 3, patient 1). Some information on false lateralization of IRSA comes from only two studies. In the first, Panet-Raymond and Gotman (15), by using power spectra, found delta activity asymmetries contralateral to epileptogenic focus in three (15%) of their 20 patients with this rhythm. In the second, Blume et al. (16) showed that correct prediction of the side of ictal onset depends on the consistency of this rhythm over several recordings. In their second group of 156 patients they found that five (4.5%) of 111 patients with a lateralized delta focus had all their seizures originating contralateral to the delta focus. One of these patients had a delta focus in two EEGs, and the other four, in single recordings, whereas none of those patients with delta foci present in three or more recordings had ictal onset contralateral to this pattern. Although based on fewer samples, our results are similar to theirs and—taken together—they suggest that IRSA may be false lateralizing in a small, perhaps ∼5%, percentage of patients with medically intractable TLE.
IRSA was almost invariably recorded from the scalp areas overlying the epileptogenic focus (temporal electrodes), and this is in line with previous observations (12,14–16,19). The topography of the interictal spikes on the scalp was similar (in the sense of greatest amplitude in the temporal electrodes), and no difference was noted among the three groups (Table 2). Such spatial concordance also is present on electrocorticographic recordings (51) and allows the hypothesis that regional spiking and slowing originate from the same region.
Topography of IRSA and spiking in children
In our 13 children, most of whom had temporal tumors, the topography of the IRSA was also temporal but wider than that in adults with similar lesions. The spatial distribution of the interictal spikes also was wider, with five (38%) of 13 children showing hemispheral or diffuse bilateral spiking. As no correlation between topography of EEG abnormalities (focal vs. diffuse) and outcome was found, these observations are in line with the notion that diffuse EEG abnormalities in children with focal lesions reflect an extended functional cortical instability, presumably related to a particular stage of brain maturation, rather than indicating multiple epileptogenic foci (52).
Inherent variability of interictal temporal slow (delta) activity and its clinical significance
Temporal slow activity has been assigned as delta by most investigators (14–16,18,19). Our results show that lateralized, primarily theta activity occurs rarely (4.5%, Table 2), and when it does so, it tends to slow to delta frequency range (of the same topography) during drowsiness and sleep. Furthermore, when theta activity is present in significant amounts, it is usually intermixed with slower rhythms. These observations suggest that, although no evidence exists that theta and delta rhythms are of different biologic significance or clinical value, delta is overwhelmingly more common, and thus the term “interictal temporal delta activity” is justified. The findings of Gambardella et al. (17), who also distinguished between theta and delta rhythms, indicated that the lateralizing value of the former frequency is clearly limited, but they assessed patients with medial temporal atrophy only and did not use additional sleep recordings to check for possible transition of theta into delta activity during sleep.
We found no convincing evidence to substantiate a meaningful differentiation between monomorphic (rhythmic) and polymorphic (arrhythmic) delta activity. Defined as “regular in shape with fairly constant duration and predominant frequency,” predominantly monomorphic delta was noted in only five (6%) of 86 patients with lateralized slow activity. In addition, monomorphic delta stretches were noted in another 27 (30%) patients, either intermixed with stretches of clearly polymorphic delta within the same run, or alternating with runs of polymorphic delta recorded by the same (temporal) electrodes (Table 1). We also observed the transition of pure polymorphic delta during awake to clearly monomorphic delta during sleep in two patients and the exactly opposite phenomenon (from monomorphic delta during awake to polymorphic delta during sleep) in another two patients. Mixed runs and coexisting rhythmic and arrhythmic patterns in the same patient render any attempt for objective classification of delta activity into distinct morphologic variants difficult or even impossible, and this might explain the discrepancy in the EEG terminology used by different authors (22), and by implication, some uncertainty with regard to interpretation of the relevant findings. For example, Reiher et al. (14) found “temporal intermittent rhythmic delta activity (TIRDA) in the one third (45 of 127) of patients with the clinical diagnosis of “complex partial epilepsy” and correctly singled it out as an accurate interictal indicator of focal epileptogenesis but did not refer to any polymorphic variant; it remains therefore uncertain whether the polymorphic pattern was not present at all, or it was recognized but not taken into account. Geyer et al. (19) distinguished rhythmic (TIRDA) and arrhythmic (TIPDA, temporal intermittent polymorphic delta activity) subtypes and showed that although the former strongly suggests temporal epileptogenesis, the latter is a poorly localizing but still an accurate lateralizing indicator. Di Gennaro et al. (53) found TIRDA in 40.3% of their patients with TLE and correlated their findings with clinical and imaging data but not with pathology. Their percentage of TIRDA is similar to ours (6% with pure TIRDA and 30% with mixed rhythmic/arrhythmic delta), but they did not consider polymorphic delta. In contrast, other authors recognized only a polymorphic form, or at least they used only the term arrhythmic or polymorphic (15–18,37,54). In our material, some runs of temporal delta were admittedly more rhythmic (or less polymorphic) than others, but all forms seemed to belong to the same continuum, the two extremes of which should perhaps be conceived as the two aspects of the same coin. One could possibly speculate that less arrhythmic patterns may reflect more intense underlying epileptic activity, and even subclinical ictal discharges; and such an assumption might well explain the findings of Geyer et al. (19). Before reaching such a conclusion, however, one has to present strong evidence that interictal temporal delta activity—at least under certain circumstances—may represent a distinct form of epileptiform activity. Finally another argument against the use of the term TIRDA is its possible confusion with rhyming and well-recognized rhythmic patterns of paroxysmal (sometimes asymmetrical and even unilateral) delta activity that, although recognized as abnormal, are neither epileptiform nor predictive of epilepsy. These are the two forms of the IRDA (intermittent rhythmic delta activity): the FIRDA (with frontal accentuation, mainly in adults and of nonspecific diagnostic relevance) and the OIRDA (with occipital emphasis, mainly in children and related to maturation rather than to some specific pathology) (55).
Abundance and reactivity
Interictal temporal delta activity was more abundant and less reactive to eye opening in the tumor group. This is certainly in line with the concept of anatomic deafferentiation of the overlying cortex (4,5), and the classic observations that polymorphic delta activity is in these cases rather continuous and shows little or virtually no reactivity (56,57). Careful analysis, however, reveals that these properties do not necessarily indicate anatomic disturbance or loss of the subcortical neuronal connections, as they were present in patients without evidence of specific structural lesion (NS group). In particular, temporal delta activity was present for 50% of the recording time in one third and was not reactive in two thirds of the patients with NS pathology, whereas some kind of reactivity (full or partial) was noted in 17.5% of patients with tumors (Table 2). Furthermore, the abundance of temporal delta activity was almost identical in MTS and NS groups. One third of patients from each group had temporal delta activity for more than half of the recording time; the rest had the rhythm for 25 to 50% of the time, and for <25% of the time in almost equal proportions in both groups (Table 2). On the assumption that anatomic neocortical temporal deafferentiation is minimal or nonexistent in patients with NS pathologic changes, these findings suggest that in the MTS group, the presence of the temporal delta activity is not significantly related to the hippocampal cell loss and the secondary degeneration of the medial-neocortical circuits. It follows that the significantly greater quantity of temporal delta activity in the patients with tumours should reflect a superimposed only element of anatomic deafferentiation.
Effects of drowsiness and light sleep
Both lateralized interictal temporal delta activity and spiking increased in quantity or appeared for the first time during drowsiness and light sleep in almost equal proportions of patients (51% for the slow and 44% for the spikes) and in all three groups. Whereas sleep-induced spike activation has been well documented (58), this is the first clear account of focal delta activation during light sleep. This similar behavior of spikes and slow waves increases the diagnostic information derived from sleep EEG studies. The affinity between temporal slow waves and spiking is further emphasized by the occasional transition of slow waves into spikes of the same topography, whereas the transformation of polymorphic forms into more monomorphic and vice versa underlines the morphologic continuity of the delta activity. Gibbs and Gibbs (59) also noted, “some slow wave foci decrease with drowsiness while others became more evident. In some cases the frequency of the abnormal rhythms changes during sleep. In other cases, epileptiform activity developed.”
Occurrence of interictal temporal delta activity on the side contralateral to tumours
Temporal slow activity was noted on the opposite side in 10 (22%) of the 46 patients with temporal lobe tumors (Fig. 6; Tables 2 and 3). Two (one with ganglioglioma, and one with DNET) had lateralized temporal slow activity contralateral to the tumor, and eight others (one with ganglioglioma, and seven with DNET) showed independent nonlateralized bilateral temporal slow activity (Table 3). Similarly, Raymond et al. (60) reported that three (19%) of their 16 patients with unilateral temporal DNET had bitemporal slow activity, which was more pronounced contralaterally in one (their case 15). However, they did not describe the relevant electrographic characteristics, nor did they comment on this paradoxical phenomenon. Blume et al. (61) reported diffuse delta activity with predominance shifting from one hemisphere to the other on sequential recordings in one of their 10 patients with tumors and persistent delta waves. Similarly, Spencer et al. (62) reported bilateral slow activity in two of 19 patients with cerebral neoplasms and in one of eight patients with nonneoplastic masses. Certainly, the presence of contralateral delta activity in patients with unilateral temporal tumors and CPSs cannot be explained by the concept of cortical deafferentiation; instead, it would suggest that this contralateral appearance is independent of the tumor.
Prognostic significance of interictal temporal slow activity
We found that in patients with minor NS pathologic changes (and by implication, with little chance to detect a relevant structural abnormality with state-of-the-art MRI technology), interictal temporal delta activity was associated with favorable outcome when it was lateralized to the side of the subsequent operation. This effect was independent of the presence or absence and the topography of interictal spikes and was significant only in the NS group. In the other two groups and in the overall patients, we observed a tendency for a favorable outcome when interictal delta and spikes were present in the later-resected temporal lobe, but this did not reach a 5% level. The reason for the interictal delta activity having a significant bearing on outcome only in the NS group is uncertain. It is likely that a similar effect in the MTS and the tumor groups might have been blurred by the generally favorable outcome and the resulting grossly uneven distribution of patients among outcome groups. This might have occurred in the previous study of Blume et al. (29). Outcome was favorable in more than four fifths of their patients, as it was in more than two thirds of all our patients, three fourths of our patients with MTS, and almost four fifths of those with tumors, but only in 37.5% of those with NS changes.
The significance of this finding becomes clear when one considers that the nature of the pathology, as is suggested by the MRI, is an important predictor of postoperative seizure control; patients with foreign tissue lesions have the best chance for a good outcome, and those with normal MRI, the worse (63), whereas the presence of hippocampal formation atrophy is associated with improvement in seizure control after temporal lobectomy, either on its own right (64) or in association with concordant epileptiform discharges (28). The high sensitivity and negative predictive value of lateralized temporal delta activity suggest an analogy to the effect of imaging and pathologic evidence of hippocampal abnormalities on outcome prediction. The absence of lateralized temporal delta may be more important as a predictive measure for a poor outcome than its presence might predict a favorable one. Such alternative interpretation also might explain why the NS group was the only one to have the strongest statistical association with interictal temporal delta activity. Given the partial overlap between MRI and temporal delta (37.5% of NS patients with delta but normal MRI as an extension of negative pathology), our observations indicate that surgical candidates with normal MRI but lateralized interictal temporal delta activity should not be barred from further assessment with ictal intracranial recordings.
Our findings suggest that interictal temporal delta activity in TLE is closely related to focal epileptogenesis, rather than being merely a product of cortical deafferentiation, and in this sense, the widely used term “background abnormality” should be restricted to alterations in basic alpha and beta rhythms in any head region, as recommended more than three decades ago (2). The pathophysiologic mechanisms underlying the appearance of interictal temporal delta activity on the scalp EEG remain uncertain. Ajmone-Marsan (65) commented on the dissociation between interictal slow activity and underlying structural change and the possible origin of the former, and emphasized that a considerable number of (deeply occurring) discharges are reflected at the scalp level in the form of irregular slow transients or, if the spikes were firing at close intervals, as trains of quasi-rhythmical slow waves. This latter situation results in a pattern at the scalp level that not only has apparently no specificity for a seizure disorder but that could also suggest an underlying structural pathology that in reality might not exist.
Conversely, one can certainly record focal delta activity with subdural electrodes from both lateral neocortical and medial temporal areas. Further studies with intracranial recordings are needed to clarify the topographic relation of the interictal delta activity to spiking and also to the area of the ictal onset.
- 3Use of EEG for evaluation of focal intracranial lesions. In: KlassDW, DalyDD, eds. Current practice of clinical electroencephalography. New York : Raven Press, 1979: 308–40..
- 4Interrelationship between cortex and subcortical structures: clinical electroencephalographic studies. Electroencephalogr Clin Neurophysiol 1953;suppl 4: 168–88., .
- 11General principles. In: EngelJJr, ed. Surgical treatment of the epilepsies. New York : Raven Press, 1993: 137–53., , .
- 30Surgically remediable temporal lobe syndromes. In: EngelJJr, ed. Surgical treatment of the epilepsies. New York : Raven Press, 1993: 49–63., , , et al.
- 31Complications of epilepsy surgery. In: EngelJJr, ed. Surgical treatment of the epilepsies. New York : Raven Press, 1993: 565–81., , , et al.
- 33An orderly approach to visual analysis: characteristics of the normal EEG of adults and children. In: DaleyDD, PedleyTA, eds. Clinical practice of clinical electroencephalography. Philadelphia : Lippincott-Raven, 1997: 139–99..
- 35The EEG and psychology of the child. In: LayryGC, ed. Handbook of electroencephalography and clinical neurophysiology. Vol 6: the normal EEG throughout life. Part B: the evolution of the EEG from birth to adulthood. Amsterdam : Elsevier Scientific, 1975: 69–104., .
- 37An orderly approach to the abnormal EEG. In: DalyDD, PedleyTA, eds. Clinical practice of clinical electroencephalography. Philadelphia : Lippincott-Raven, 1997: 253–67., .
- 40The neuropathology of temporal lobe epilepsy. New York : Oxford University Press, 1988..
- 42Histological typing of tumors of the central nervous system. 2nd ed. Heidelberg : Springer-Verlag, 1993., , .
- 43The neuropathology of temporal lobe epilepsy. Mod Trends Neurol 1970;5: 254–70..
- 45Outcome in respect to epileptic seizures. In: EngelJJr, ed. Surgical treatment of the epilepsies. New York : Raven Press, 1993: 609–21., , , et al.
- 47Abnormalities of gyration, heterotopias, tuberous sclerosis, focal cortical dysplasia, microdysgenesis, dysembryoplastic neuroepithelial tumour and dysgenesis of the archicortex in epilepsy: clinical, EEG and neuroimaging features in 100 adult patients. Brain 1995;118: 629–60., , , et al.
- 52Extracranial EEG evaluation. In: EngelJJr, ed. Surgical treatment of the epilepsies. New York : Raven Press, 1993: 173–95., , .
- 54Focal brain lesions. In: DalyDD, PedleyTA, eds. Current practice of clinical electroencephalography. Philadelphia : Lippincott-Raven, 1997: 335–70., .
- 56Brain tumors and other space-occupying lesions. In: NiedermeyerE, Lopez da SilvaFH, eds. Electroencephalography: basic principles, clinical applications, and related fields. Baltimore : Williams & Wilkins, 1993: 268..
- 59Atlas of electroencephalography. Vol Three: neurological and psychiatric disorders. Reading , Mass : Addison-Wesley, 1964., .