24-Hour Video EEG in the Evaluation of the Seizure-Free Patient Before Antiseizure Medication Withdrawal

Abstract

Purpose:

In patients with epilepsy who achieve seizure freedom, physicians may consider discontinuing antiseizure medications, often using EEG beforehand to guide the decision. This study evaluates the gain in detection of epileptiform discharges (EDs) in 24-hour video EEG (VEEG) monitoring in seizure-free patients.

Methods:

The authors identified patients with epilepsy who were seizure free on antiseizure medications and had undergone 24-hour VEEG after an unrevealing routine EEG. The authors evaluated the yield and latency of observed EDs in the VEEG study. The authors compared the rate of ED detection during the first 60 minutes versus later in the recording.

Results:

Of the 27 patients, aged 19 to 51 years, 10 (37%) exhibited EDs on 24-hour VEEG. The latency to the first EDs ranged from 52 to 748 minutes, with a median of 164 minutes. Nine of these 10 patients (90%) had their EDs appear only after the first 60 minutes of recording. In other words, prolonging the EEG beyond 1 hour yielded an additional 33.3% of patients with EDs that a 1-hour recording would have missed. Younger age and presence of EDs on a past EEG were predictive of the presence of EDs on VEEG (P = 0.040 and P = 0.039 respectively).

Conclusions:

The results suggest that 24-hour VEEG is more sensitive in detecting EDs than routine EEG in seizure-free patients and thus may be superior in individualized risk assessments for seizure recurrence on potential antiseizure medication discontinuation.

Epilepsy is one of the most common neurologic diagnoses worldwide and is mainly treated with antiseizure medications (ASMs) on a chronic basis.^1,2 Although they lead to seizure freedom in approximately 70% of patients,^2–4 ASMs are often a source of inconvenience because of significant economic and psychosocial burdens, concerns for teratogenicity, and other prominent, unpleasant systemic side effects and neurologic side effects resulting from central nervous system modulation.^2,5 Thus, once complete seizure freedom on ASMs is achieved and sustained, patients and physicians often discuss whether medications can be safely discontinued. However, the decision to taper and withdraw ASMs in a seizure-free patient is complex, because the potential risk of relapse can exceed 30%.^6,7 Seizure recurrence on medication discontinuation poses several mental and physical risks, and concerns for safety, especially in everyday situations such as driving or working.^8–10

With the absence of clear-cut clinical guidelines for withdrawing ASMs, physicians often encounter a dilemma with this decision, relying on a complex process of balancing potential gains and acceptable risks.^8–10 Hence, the management is tailored for each subject and often involves shared decision making between the patients and their physicians. Several independent risk factors have been proposed as predictors for seizure recurrence in patients after withdrawing ASMs, including the presence of epileptiform discharges (EDs) on EEG.^7,8,11–13

The weight to be allocated for the EEG abnormalities predictive value is a matter of debate. Some authors argue that such findings, if isolated, should not necessarily preclude the withdrawal of an ASM. In this specific context, the EEG literature used to quantify its predictive value does not specify what kind of EEG was performed, and does not typically include prolonged recordings.^10,12,14,15

The sensitivity to detect EDs in patients with active epilepsy is low with single routine EEGs but has been proven to be enhanced by performing 24-hour (or longer) EEG studies.^16–20 The same has also been shown in the context of a first seizure.^21,22 Such information is lacking in the context of epilepsy with sustained seizure freedom; a practice advisory published on behalf of the American Academy of Neurology Guideline Subcommittee in 2021 concluded that they were unable to determine the length of EEG needed to assess the risk of seizure recurrence after ASM withdrawal because of lack of specific studies exploring this question.¹⁵ Based on the assumption that short routine EEGs may be insufficient to be incorporated in the stratification of the risk of seizure recurrence in well-controlled patients, some are now being referred for 24-hour EEG recordings before the consideration of ASMs discontinuation. This article attempts to evaluate this practice at the center in terms of gained yield in EDs detection, specifically in seizure-free subjects.

METHODS

Study Design and Participants

The study was approved by the institutional review boards at Hamad Medical Corporation and Weill Cornell Medicine-Qatar under the assigned IRB number 23-00010. Because the data were collected retrospectively, informed consent was waived.

Suitable study subjects were identified from the adult epilepsy monitoring unit admissions log at Hamad Medical Corporation. All subjects fulfilling the inclusion and exclusion criteria and monitored between June 2019 and June 2023 were included. Inclusion criteria consisted of a history of epilepsy, seizure freedom for at least 2 years with the usage of one or more ASMs, and documentation of a 24-hour Video EEG (VEEG) in the work-up before potential discontinuation of ASMs. All patients lacked EDs on their most recent documented routine EEG before undergoing VEEG. Exclusion criteria included patients with other neurologic disorders mimicking epilepsy, patients not receiving ASMs, and patients who have undergone surgical epilepsy treatment. Extracted data included patient sex, age, age at seizure onset, epilepsy risk factors, seizure type(s), epilepsy syndrome(s), length of seizure freedom, past investigations, type of ASM, and the results of both the patients past routine EEGs and the more recent VEEG data. The collected data were deidentified and tabulated.

EEG: Measures and Outcomes

All study subjects encountered a 24-hour VEEG recording once, typically weeks to months after an unrevealing routine EEG. Video EEG data were recorded by using the standard 10-20 system for electrode placement. The first 60 minutes of the recordings were acquired as per our extended EEG protocol and included activation procedures such as hyperventilation, photic stimulation, and attempted sleep. The VEEG data were reviewed by electroencephalographers at Hamad Medical Corporation and classified as normal or abnormal, with specific attention to the appearance of EDs, their distribution, latency, and the behavioral states they appear in (wakefulness, sleep, or both). Epileptiform discharges were defined as spikes, sharp waves, and spike or polyspike and wave discharges. Nonepileptiform abnormalities, such as background slowing, were also noted. Data pertaining to whether EDs were previously found in past EEGs for each patient were also extracted from the records. Cumulative sensitivity of the prolonged 24-hour VEEG data was assessed by recording the latencies of EDs when present and specifying whether they were observed within the first hour of the recording or later.

Statistical Analyses

Continuous variables were summarized using means ± standard deviations along with medians and interquartile ranges. Categorical variables were summarized using frequency distributions. Comparative analyses were conducted to examine the relationship between VEEG and clinical variables. The Mann–Whitney U test was used to assess the relationship between the VEEG results and continuous variables such as age, age at first unprovoked seizure, and number of years of seizure freedom. Fisher exact test was used to evaluate the relationship between VEEG epileptiform abnormalities and several categorical variables, such as sex, number of ASMs tried, total number of seizures, epilepsy classification, behavioral state during which seizures occurred (sleep and/or wakefulness), family history, imaging, and early EEG findings. To investigate the diagnostic yield of prolonged VEEG compared with the initial 60 minutes of recording, relative risk reduction and absolute risk reduction (RRR and ARR) calculations were estimated. Using the ARR difference, the number needed to treat was calculated to quantify the number of patients who would benefit from undergoing 24-hour VEEG to observe an increased yield relative to routine EEG before discontinuing ASMs. All data analyses were performed using IBM-SPSS (version 29, Armonk, NY). Statistical significance was set at the 5% level.

RESULTS

Data on 27 patients were collected in this study, including 15 women (56%) and 12 men (44%). Patients' age ranged from 19 to 51 years. All patients were seizure free for at least 2 years, with a median of 4 years. Nine patients (33%) had >10 seizures during their lifetime. All 27 subjects experienced generalized tonic–clonic seizures by description, with the latter including also the focal to bilateral tonic–clonic events. Eighteen of the 27 (67%) experienced solely generalized tonic–clonic seizures and the others had either other focal, absence, and/or myoclonus seizures alongside generalized tonic–clonic seizures. There were no cases of status epilepticus. Thirteen patients (48%) had generalized epilepsy, nine (33%) had focal epilepsy, and the remaining five (19%) had an epilepsy type that was unclassified as either focal or generalized in origin. All patients had grossly normal neurologic examination. The subjects were on a wide variety of medications, including levetiracetam, valproic acid, lamotrigine, topiramate, lacosamide, and carbamazepine; all patients were on monotherapy at the time of the VEEG except one, who was on two ASMs.

All patients had no epileptiform activity in their last documented EEG before the VEEG but nine (33%) demonstrated epileptiform activity in their past EEG on initial diagnosis and remote work-up. Four patients (15%) had abnormal imaging findings with a likely epileptogenic structural pathology (namely cortical/subcortical focal encephalomalacia or gliosis, post-trauma, or postsurgery). An additional three patients (11%) had abnormal imaging with nonspecific changes (white matter microangiopathic ischemic changes or global atrophy). Individual patient demographics, seizure characteristics, and EEG findings are portrayed in Table 1.

TABLE 1.

Individual Demographics, Seizure, and EEG Characteristics

Patient	Age (Years)	Age at First Unprovoked Seizure (Years)	Seizure Type(s)	Length of Seizure Freedom (Years)	Epilepsy Type (Syndrome)	Initial EEG EDs	VEEG EDs	VEEG Latency (Minutes)
1	38	35	BTC	3.5	Focal	Yes	Yes, focal extratemporal	748
2	43	23	BTC	8	Unclassified	No	No	—
3	48	30	BTC	5	Focal	No	No	—
4	32	28	BTC	3	Unclassified	No	No	—
5	24	20	GTC	2.6	Generalized	No	No	—
6	27	8	GTC + myoclonus	6.5	Generalized (JME)	No	No	—
7	19	10	GTC	5	Generalized	No	Yes, generalized	155
8	44	25	GTC + myoclonus	6	Generalized (JME)	No	No	—
9	30	9	BTC	10	Focal	No	No	—
10	44	41	BTC	2	Focal	No	No	—
11	20	12	GTC	5	Generalized	No	Yes, generalized	97
12	20	15	BTC + FA	3	Focal	No	No	—
13	23	18	BTC	5	Focal	Yes	No	—
14	37	32	BTC + FIA	2	Focal	No	Yes, focal temporal	52
15	22	16	GTC	5	Generalized	Yes	No	—
16	31	14	GTC	3	Generalized	No	No	—
17	21	11	BTC	3	Focal	No	No	—
18	35	32	BTC	3	Unclassified	No	No	—
19	20	17	GTC + absence	3	Generalized (JAE)	Yes	Yes, generalized	248
20	24	16	GTC + myoclonus	4	Generalized (JME)	Yes	Yes, generalized	142
21	24	7	GTC + absence	6	Generalized (CAE)	Yes	No	—
22	28	18	GTC	3	Generalized	No	Yes, generalized	173
23	51	43	BTC	6	Unclassified	No	No	—
24	25	18	GTC + myoclonus	5	Generalized (JME)	Yes	Yes, generalized	226
25	20	15	BTC	5	Focal	Yes	Yes, focal temporal	83
26	20	15	GTC + absence	3	Generalized (JAE)	Yes	Yes, generalized	380
27	28	20	BTC	3.5	Unclassified	No	No	—

BTC, bilateral tonic–clonic seizure; CAE, childhood absence epilepsy; EDs, epileptiform discharges; FA, focal aware seizure; FIA, focal impaired awareness seizure; GTC, generalized tonic–clonic seizure; JAE, juvenile absence epilepsy; JME, juvenile myoclonic epilepsy; VEEG, video-EEG.

Of the 27 patients, 10 (37%) had EDs on the VEEG study, of whom 7 had generalized and 3 had focal activity in nature. This meant that such EDs would have been missed in more than one-third of subjects had the authors solely relied on the results of the latest routine EEG. Five patients (19%) had nonspecific abnormalities (focal or generalized slowing). Of the 10 subjects with EDs, 6 manifested this activity only during sleep, while 4 had the EDs both during sleep and wakefulness. The relationship of several categorical variables with VEEG epileptiform abnormalities is displayed in Table 2. Statistical significance was only observed for comparison between initial EEG epileptiform activity and VEEG abnormalities; subjects with initial EEG EDs had a higher incidence of EDs on the VEEG than those without (66.7 vs. 22.2%, P = 0.039).

TABLE 2.

Subgroup Breakdowns of Categorical Variables of Interest and Their Relationship With VEEG EDs Portrayed With P-values

		VEEG Epileptiform Abnormalities				P
		No		Yes
		Count	%	Count	%
Sex	Female	10	66.7	5	33.3	0.706
	Male	7	58.3	5	41.7
Number of ASMs tried	1 ASM	9	60	6	40	>0.999
	>1 ASM	7	63.6	4	36.4
Total number of seizures	<10	10	58.8	7	41.2	0.418
	>10	7	77.8	2	22.2
Behavioral state at seizures	Asleep	2	50	2	50	0.092
	Awake	4	44.4	5	55.6
	Both	7	100	0	0
	Unknown	4	57.1	3	42.8
Family history of seizures	Absent	12	66.7	6	33.3	0.683
	Present	5	55.6	4	44.4
Imaging	Normal	11	55	9	45	0.204
	Abnormal	6	85.7	1	14.3
Initial EEG EDs	No	14	77.8	4	22.2	0.039
	Yes	3	33.3	6	66.7
Epilepsy classification	Focal	6	66.7	3	33.3	0.127
	Generalized	6	46.2	7	53.8
	Unclassified	5	100	0	0

ASMs, antiseizure medications; EDs, epileptiform discharges; VEEG, video-EEG.

The bold P-values of less than 0.05, indicate statistical significance.

The summary of several numerical variables is highlighted in Tables 3 and 4, along with the relationship with VEEG epileptiform abnormalities. Statistical significance was only achieved by comparison between patient age and VEEG epileptiform activity (P = 0.040), with a younger age predicting a greater likelihood of epileptiform activity.

TABLE 3.

Numerical Data of Interest Summarized With Mean, Median, and Standard Deviation

	Mean	SD	Median
Age (years)	29.56	9.66	27
Age at first unprovoked seizure (years)	20.3	9.85	18
Duration of epilepsy (years)	6.51	5.46	4
Number of years of seizure freedom	4.42	1.87	4
VEEG EDs latency (minutes)	230.4	205.15	164

EDs, epileptiform discharges; VEEG, video-EEG.

TABLE 4.

Relationship of Numerical Variables of Interest (Means, Medians, SDs, IQRs) With VEEG EDs Highlighted Using P-values

	VEEG Epileptiform Abnormalities				P
	No		Yes
	Mean (SD)	Median (IQR)	Mean (SD)	Median (IQR)
Age (years)	32.2 (10.2)	30.0 (23.5–43.5)	25.1 (7.1)	22.0 (20.0–30.3)	0.040
Age at first unprovoked seizure (years)	21.2 (10.9)	20.0 (12.5–29.0)	18.8 (8.2)	16.5 (14.3–21.5)	0.675
Number of years of seizure freedom	4.7 (2.2)	5.0 (3.0–6.0)	3.9 (1.1)	3.8 (3.0–5.0)	0.386

EDs, epileptiform discharges; IQR, interquartile range; VEEG, video-EEG.

The bold P-values of less than 0.05, indicate statistical significance.

The time taken for EDs to first manifest during VEEG recording varied from 52 to 748 minutes, with a median of 164 minutes. Among the 10 patients exhibiting epileptiform activity on VEEG, 9 experienced them only after the 60-minute mark of recording, a time frame well beyond what routine and extended EEGs typically cover. This marks a substantial 90% increase in relative yield and a 33.33% increase in absolute yield and indicates that most of these EDs would have gone unnoticed had VEEG been substituted with either a repeat routine or even extended EEG recording of <60 minutes duration. In practical terms, for each three seizure-free patients undergoing 24-hour EEG, one additional patient was identified with EDs that would have been otherwise missed with just a 60-minute EEG instead.

DISCUSSION

Two-thirds of patients with epilepsy (PWE) achieve seizure freedom with appropriate ASMs.⁴ When such seizure control is maintained, the dilemma of continuing the drugs or not arises. Self-initiated discontinuation of ASM treatment without appropriate medical advice is not uncommon in PWE. In addition to the high risk of seizure recurrence and the need to restart treatment, 28% experienced a seizure-related injury after treatment self-discontinuation.²³ This highlights the need for improved discussion between treating physicians and patients on this specific issue.²⁴ By discontinuing ASMs, side effects, drug interactions, teratogenicity, costs, and stigma can be avoided. However, this comes at the expense of higher risk for seizure recurrence with all its consequences (injury, death, occupational, and driving restrictions).^8,10,11 Many individual risk factors for seizure relapse have been identified, including the presence of paroxysmal EEG abnormalities.¹² Prediction models, based on the cumulative risk from multiple variables including EEG findings, can better assist in quantifying the risk of seizures after ASM withdrawal.^7,25 The 1996 American Academy of Neurology practice parameter suggested that the discontinuation of ASMs may be considered if the patient meets a certain profile that includes a normalized EEG.²⁶ The sensitivity to detect EDs in EEG increases with the duration of the recording in patients with active epilepsy and in the context of a first unprovoked seizure.^16–22 The superiority of prolonged recordings is less well established in PWE who become seizure free; a practice advisory update by the Academy of Neurology could not determine the optimal length of EEG to assess the risk of seizure recurrence when ASM withdrawal is considered in seizure-free patients, because studies addressing the predictive value of EEG in this specific scenario did not specify the duration of recordings and mostly relied on routine studies.¹⁵ The authors’ goal is to address this specific question: in seizure-free patients, is 24 hour EEG superior to shorter single or repeated recordings?

In this study of 27 seizure-free patients with normal most recent routine EEG, the 24-hour recording revealed new EDs in 10 patients. This corresponds to an absolute increase of 37% in ED detection rate and in practical terms means that more than one of each three patients would have been falsely labeled as having a normalized EEG had the authors relied solely on the routine EEG, with a potentially higher risk than anticipated for seizure recurrence on medication withdrawal. Furthermore, most of these EDs (9 out of 10) were seen only after the 60-minute initial portion of the recording (which ran according to the extended EEG protocol with the typical activation procedures and an attempt to sleep). Thus, had the authors substituted the 24-hour EEG with just another routine or extended EEG, most of the gained yield would have been erased.

The literature on the usage of 24-hour EEG specifically before ASM withdrawal is limited. One study that investigated the usage of long-term ambulatory EEG in ASM withdrawal for PWE in long-term remission found that abnormal results predicted seizure recurrence; however, this study only included 15 institutionalized adults who had learning disabilities and cognitive impairment in addition to epilepsy.²⁷ Another study investigated the yield of long-term electroencephalography as compared with routine EEG, and found that the longer studies doubled the yield for EDs.²⁸ The study only included patients with focal epilepsy of unknown cause and excluded those with generalized epilepsy. A more recent study concluded also that long-term electroencephalography had improved sensitivity; however, the recording varied in length, and could have been as short as 3 hours. The patient population in this study represented a heterogeneous group with several patients having multiple neurologic deficits, limiting its generalizability. The study also did not assess the latencies at which EDs occurred.²⁹ This study sample is more representative of the general drug responsive epilepsy population, without overt neurologic deficits.

The increased sensitivity of 24-hour VEEG is most likely due to improved sampling because EDs occur sporadically and may not be detected in a single 1-hour window. In addition, circadian rhythms and sleep-wake patterns have long been known to influence the frequency of EDs that may be exclusively activated during sleep.³⁰ This was observed in this study, because all 10 patients with EDs on VEEG experienced them during sleep, of which 6 exclusively had EDs during sleep and not in the awake state; no patient experienced EDs only while awake. Furthermore, one meta-analysis found that EDs occurred 2.46 times more frequently in stage 3 of Non-REM (NREM) sleep as compared with REM sleep, which supports the importance of evaluating a full cycle of sleep during an EEG recording to observe the most fruitful period of EDs in the circadian rhythm.³¹ In addition, longer recordings by nature provide more data for readers to compare distinct waveforms and EDs to one another, thus increasing the confidence in determining whether a wave pattern has an epileptiform morphology. This can help electroencephalographers distinguish EDs more reliably from artifacts or benign sharply contoured activity.

The authors found two predictors of the presence of EDs on the 24-hour VEEG: younger age and the past documentation of such discharges in a remote EEG. Patients with EDs were on average 7 to 8 years younger than those without. This is consistent with existing literature that describes the lower detection of EDs in older PWE, possibly because of an age-related decrease in the complexity of synaptic connections leading to less frequent synchronized neuronal discharges.³² Epileptiform discharges also change in morphology with age, often having lower amplitudes and less obvious qualities of sharp and slow patterns.³³ The detection of EDs on an early EEG is predictive of their continued presence on the 24-hour VEEG in seizure-free PWE, even with reported normalized routine EEGs in between. To the authors’ knowledge, this correlation in this specific context has not been reported in the literature previously.

The detection of EDs on VEEG in this study did not significantly differ between patients with other characteristics, including sex, age at seizure onset, family history of seizures, number of seizures, occurrence of seizures in wakefulness or sleep states, length of seizure freedom, or the number of tried ASMs. This lack of significant associations between some patient characteristics and VEEG findings should be interpreted cautiously because of the limited sample size in this study.

Although focal epilepsies are more common than generalized epilepsies in the general population, almost half of this study cohort featured patients with idiopathic generalized epilepsies. The overrepresentation of such syndromes in this seizure-free population contemplating discontinuation of ASMs is in line with previous observations that revealed a higher proportion of seizure freedom on ASMs in patients with generalized idiopathic or genetic epilepsies.^34,35

The relatively small sample poses limitations to the generalizability of the findings. In addition, the authors lacked long-term outcomes as to whether the patients did indeed stop their ASMs or experienced seizures if they have done so. This is planned in a future larger and extended study. These VEEG studies were performed on an inpatient observation basis because the authors lack prolonged ambulatory EEG capabilities. Continuous ambulatory EEG represents an adequate and cheaper alternative with a likely similar high yield to inpatient VEEG, in the assessment before ASM withdrawal.

In conclusion, this study suggests that 24-hour EEGs detect more EDs than traditional shorter recordings in seizure-free patients contemplating ASM discontinuation. Such longer recordings also seem to be a better alternative than just performing serial and repetitive routine/extended EEGs. This finding calls on the re-evaluation of the predictive value of EEGs in this context, with the use of the current widely available capabilities for prolonged recordings, including in the ambulatory setting, coupled with longer clinical follow-up. Larger prospective studies would be needed to achieve this goal. If 24-hour EEG results (rather than short-recording EEG data from older literature) are used to re-establish an individual's recurrence risk after medication withdrawal, then the weight given to EEG findings in predictive models should be re-evaluated and updated.