Fig. 1
Extreme delta brush pattern (arrow) and seizure onset in a patient with NMDA encephalitis
Patients with antibodies to the leucine-rich glioma-inactivated 1 (LGI1) protein component of the voltage-gated potassium channel (VGKC) have typical limbic encephalitis symptoms and develop memory disturbances, confusion, and seizures. Memory and cognitive deficits may be preceded by short faciobrachial dystonic seizures. These seizures often do not respond to AED therapy. Patients may develop hyponatremia or rapid eye movement sleep behavior disorders. Only 20 % of cases are paraneoplastic; the most common associated tumors are thymoma or lung cancer. Early diagnosis is important to prevent progression from seizures to other symptoms of encephalitis and because these patients have a dramatic response to corticosteroid treatment [17].
There are several other causes of autoimmune encephalopathy in which seizures play a significant role that are not associated with underlying malignancy or limbic encephalitis. One type of presentation of Hashimoto’s encephalopathy includes seizures, acute deterioration of consciousness, and stroke-like episodes with the presence of antithyroid antibodies but with normal or slightly abnormal thyroid function. The most common seizure pattern includes generalized tonic-clonic seizures followed by complex partial seizures, with or without secondary generalization [18]. Other systemic autoimmune diseases associated with seizures that usually respond to steroid therapy include systemic lupus erythematosus, Sjogren’s syndrome, Wegener’s granulomatosis, and neurosarcoidosis [12].
Posterior Reversible Encephalopathy Syndrome
Posterior reversible encephalopathy syndrome (PRES) is characterized by altered mental status, seizures, and visual changes accompanied by characteristic neuroimaging changes in a posterior symmetrical distribution, interpreted as edema. Altered mental status, headache, and visual disturbances are the classic clinical findings and seizures are frequently reported. Most seizures are single short grand mal seizures but multiple grand mal seizure and focal seizures can sometimes occur [19]. It is associated with a variety of underlying clinical conditions that may cause seizures and encephalopathy such as electrolyte disturbances. Treatment of the underlying cause is paramount and rapid initiation of AEDs may help to prevent further neuronal injury from seizures.
Subacute Encephalopathy with Seizures in Chronic Alcoholism
Subacute encephalopathy with seizures in chronic alcoholism (SESA) is a rare clinical syndrome which describes alcoholic patients presenting with confusion, seizures, and focal neurologic deficits. The occurrence of SESA may be underestimated, as many patients may be misdiagnosed with alcohol withdrawal seizures. It has recently been proposed that this syndrome may be a type of NCSE because these patients have focal seizures between which they have PEDs and do not return to baseline mental status [20]. Therefore, cEEG monitoring is recommended for alcoholic patients with encephalopathy and focal neurologic deficits. An accurate diagnosis is critical because these patients require long-term treatment with antiepileptic medications to prevent recurrence [12].
Other Rare Causes
There are a variety of other systemic diseases that can contribute to both encephalopathy and seizures, including hepatic failure, uremia, human immunodeficiency virus (HIV) infection, and drug intoxication (both prescription as well as recreational) [12].
Clinical Significance of Seizures
The detection of seizures in ICU patients is associated with poor prognosis as defined by death or severe disability at hospital discharge [1]. It is unclear if the seizures themselves worsen prognosis or if seizures are simply found more often in patients who are more seriously ill. Therefore, it is unclear if the treatment of intermittent seizures improves prognosis. Generalized convulsive status epilepticus (GCSE), on the other hand, if not controlled almost certainly leads to irreversible neuronal injury, and therefore treatment of GCSE almost certainly improves prognosis.
If we are not sure that treatment of seizures changes prognosis, why should we monitor patients? At this point, most monitoring is done with the assumption that the treatment of seizures improves patient care. The detection of seizures in patients in the ICU certainly affects their medication management. A study of 300 consecutive monitoring studies of 287 adult and pediatric inpatients demonstrated that cEEG monitoring led to a change in AED prescribing in 52 % of cases. There was initiation of an AED in 14 %, modification of an AED regimen in 33 %, and discontinuation of AED therapy in 5 % [3]. Many of these changes were due to the detection of seizures. A similar study in children reported the CEEG monitoring affected the care of children of 59 % of cases, again mostly by affecting AED management [21].
It is important to accurately recognize electrographic seizures. There are many non-epileptic events, either clinical or electrographic, which can be difficult to distinguish from epileptic seizures on EEG or video. Treatment of non-epileptic events with AEDs can cause harm to patients by increasing sedation in an already susceptible patient population. Lorazepam, for instance, is routinely used to abort seizures and has been shown to be an independent risk factor for delirium in ICU patients. Increased risk of delirium may translate to increased morbidity and mortality, prolonged hospital stay, and increased health-care costs [22].
Electrographic Criteria for Labeling Seizures
Definition of Electrographic Seizure
Although there is reasonable general acceptance of what constitutes a clinical seizure, the variety of electrographic and clinical events observed during the care of critically ill patients make it difficult to define a specific electrographic criterion for seizures. For non-seizure periodic and paroxysmal activity, a committee of the American Clinical Neurophysiology Society (ACNS) has recently established research terminology for some of the more frequently encountered electrographic patterns for ICU monitoring based on whether they are generalized or localized and their periodicity, persistence, duration, frequency, inducibility by external stimuli, and evolution [23]. Although there is no equivalent established terminology for electrographic seizures, the EEG terminology developed for NCSE may be adapted to cover most types of seizures seen in the ICU. In this terminology [24], an event would meet criteria for a seizure if it consisted of epileptiform discharges at a frequency of greater than 2.5 Hz or if it contained epileptiform discharges or rhythmic delta/theta activity at less than 2.5 Hz accompanied by one of the following three criteria: (1) EEG and clinical improvement after an IV AED, (2) subtle clinical ictal phenomena during the EEG patterns, or (3) typical spatiotemporal evolution. Spatiotemporal evolution refers to a characteristic of most seizures whereby the EEG pattern “evolves” or changes over time in either its frequency or in its spatial distribution. In the case of a typical seizure, this usually involves a gradual increase or decrease in the frequency of the EEG rhythm accompanied by spread of the EEG activity to a greater number of EEG electrodes. The 2012 ACNS Standardized Critical Care EEG Terminology lists definite seizure activity to be generalized spike and wave patterns of a frequency greater than 3/s and evolving discharges that are of a frequency equal or greater than 4/s. Patterns that do not meet this criteria are not ruled out as seizure activity, only deserving of greater scrutiny [23]. The Pediatric Critical Care Terminology lists the minimum duration for a seizure to be 10 s, and this is a generally accepted criterion for adult seizures as well.
Borderline Seizure Patterns
Brief Ictal Rhythmic Discharges
Most electroencephalographers use a definition of seizures that includes a minimal duration of 10 s, which reflects the typical lower limit to the duration of focal seizures in patients with chronic epilepsy. Rhythmic ictal-appearing patterns lasting less than 10 s have been described in neonates under multiple different acronyms including brief ictal rhythmic discharges (BIRDs). In neonates, these patterns encompass discharges of any frequency, including less than 4 Hz, because they are common in this age group. Brief bursts of rhythmic delta activity and periodic discharges with a frequency of less than 4 Hz are common in critically ill patients but are usually not considered to be ictal. Ictal discharges in children and adults often have a higher frequency than those in neonates. The occurrence of BIRDs with a frequency higher than 4 Hz has recently been reported in 20 out of 1135 CEEG recordings [25]. The typical frequency for BIRDs was in the theta, alpha, and beta frequency bands in 14 (70 %), 3 (15 %), and 3 (15 %) cases, respectively. Typical duration was 1–3 s. Most (17 of 20 [85 %]) BIRDs were sharply contoured except in the theta frequency band in two patients (10 %) and the beta frequency band in one patient (5 %), which were sinusoidal. None of the BIRDs showed obvious evolution. Most of the patients with BIRDs had acute brain injuries such as tumor and stroke, and most patients were comatose or stuporous. BIRDs were present in the first hour of the recording in most patients and recordings with BIRDs were more likely to also contain seizures than recording without BIRDs. Figure 2 shows an example of two BIRDs in one page of EEG.
Fig. 2
Two brief ictal rhythmic discharges (BIRDs) in the left hemisphere
Lateralized Periodic Discharges
Lateralized periodic discharges (LPDs) (previously known as periodic lateralized epileptiform discharges (PLEDs) in older nomenclature [23]) are associated with nearly any type of structural abnormality including those due to infection, neoplasm, ischemia, hemorrhage, and anoxia. They are associated with poor prognosis, particularly in patients with neoplasms. They are not generally considered to indicate seizure activity, although this is debated. Evidence indicating that LPDs may indicate ongoing focal seizure include reports of altered mental status in the elderly patients associated with LPDs and increased metabolic activity present in regions with LPDs in positron emission tomography (PET) and single-photon emission computed tomography (SPECT) scans. But most LPDs are not considered seizures because many patients have LPDs chronically, and in patients with both LPDs and seizures, the seizures appear distinct from the LPDs and the LPDs stop during the seizures. Some types of LPDs are considered more likely to indicate focal seizure activity, and these have been termed “PLEDs-plus” [26] in older terminology, now probably known as “LPDs-plus.” The periodic discharges which make up the LPDs-plus pattern include brief focal rhythmic activity, and/or the patient has physical manifestations which correlate with these LPDs such as rhythmic movements or myoclonus. Figure 3 shows left central LPDs which are accompanied by right facial twitching, so they are considered LPDs-plus and therefore evidence of focal seizure activity. Figure 4 shows a LPDs-plus pattern which is not accompanied by clinical manifestation of seizure but should be considered evidence of focal seizure activity because the EEG pattern of LPDs is combined with rhythmic alpha and theta activity.
Fig. 3
Left central LPDs which are accompanied by right facial twitching, consistent with the LPDs-plus pattern and therefore evidence of focal seizure activity
Fig. 4
LPDs-plus pattern in the right temporal-parietal region which can be considered to be evidence of focal ongoing seizure activity
Stimulus-Induced Rhythmic Periodic or Ictal Discharges
Stimulus-induced rhythmic, periodic, or ictal discharges (SIRPIDs) are found in approximately 20 % of patients undergoing cEEG monitoring. These are rhythmic frontally predominant generalized periodic discharges which occur when a patient is stimulated. They are considered to fall somewhere along the ictal-interictal continuum. Clinical or subclinical/electrographic seizures are found in about half of these patients; SE is found more frequently in focal or ictal-appearing SIRPIDs [27]. As such, treatment with a conventional AED is advisable. But studies have shown no increase in regional cerebral blood flow to indicate that they may represent seizure activity, and as a result aggressive treatment is not recommended. After cardiac arrest, SIRPIDs are associated with poor outcome, especially during hypothermia, but in other instances, outcome is yet to be defined. Figure 5 shows EEG from a patient during a SIRPID episode, and Fig. 6 shows EEG from the same patient during a period without SIRPIDs.
Fig. 5
An episode of SIRPIDs, as manifested on EEG by GPDs
Predictors of Electrographic Seizures
There are multiple elements of the past medical history, neurological exam, and interictal EEG that predict the presence of electrographic seizures in cEEG recordings. Predictors in the past medical history include young age, a history of epilepsy, and remote risk factors for seizures including brain injury. Elements of the history of present illness which predict seizures are a report of convulsive seizures before the EEG recording is begun, sepsis, and recent cardiac or respiratory arrest. Findings on the neurological exam which predict seizures include oculomotor abnormalities including nystagmus, hippus, and eye deviation. Electrographic features which predict seizures include epileptiform discharges including spikes, sharp waves, LPDs, and generalized periodic discharges (GPDs) [1, 28]. Patients without epileptiform discharges in the first 30 min of an EEG recording have approximately a 10 % chance of developing seizures in the subsequent cEEG recording. If patients have epileptiform discharges in the first 30 min, the changes of recording a seizure are significantly elevated to around 25 %. Patients with no epileptiform discharges in the first 2 h of the EEG recording have less than a five percent change of developing subsequent seizures. More than 95 % of seizures are recorded in the first 24 h of EEG monitoring, so if a patient does not have a seizure in 24 h of monitoring, it is unlikely (less than 5 % chance) that seizures will be recorded with further EEG monitoring, even if epileptiform discharges are present [29].
Inter-rater Agreement for Labeling Seizures
Because the performance of automated seizure detection programs has not been verified in sizable studies, it is not consistently used in clinical practice. Therefore, seizures are usually identified by visual inspection of the unprocessed EEG recordings. This is a significant challenge because these recordings can be quite long and seizure patterns can be subtle. Inter-rater performance for experts in labeling seizures is not perfect. In a recent study of eight board-certified EEG experts who independently identified seizures and periodic discharges (PDs) in 31-h ICU EEG segments from three medical centers, the inter-rater correlation between the experts was only moderate. But the correlation of experts for labeling of seizures was considerably higher than for the labeling of periodic discharges. Improved performance in labeling seizures and PDs was seen in experts who had received specific training by the Critical Care EEG Monitoring Research Consortium [30].
Automated Detection of Seizures
The first general-purpose seizure detection methods were introduced in the 1980s. However, none of the seizure detection software currently available on the market has been shown to be as accurate as a human for reviewing long-term cEEG recordings. For example, Reveal (by Persyst Development Corporation), one of the most advanced seizure detection products, offers an average sensitivity of 85 % with a false detection rate of about 14 per day for adult epilepsy monitoring unit (EMU) patients [31]. Reveal’s false detection rate often becomes extremely high (>40/day) when an EEG recording contains high number of recording artifacts, a problem common in EMU scalp EEG monitoring and more frequently seen in ICU studies. Due to this high false detection rate, most centers do not utilize or rely on seizure detection software in the long-term EEG study review process. In addition to insufficient detection performance, another major problem is that all commercially available seizure detection software is marketed for use across all patient populations, but has not been clinically validated for each patient population. These substantially different patient populations include adult and pediatric EMU patients (with scalp and intracranial recordings), adult and pediatric in home (ambulatory) EEG patients, and neonatal, pediatric, and adult ICU patients. It has been well documented that, despite of some underlying similarities, there exist significant differences in both ictal and background EEG patterns among these patient populations. Therefore, it is very difficult, if not impossible, for one algorithm to perform well enough in all cEEG patient populations to be clinically useful. An ideal seizure detection system must include different modules for use in different patient populations, and the performance of each must be clinically validated using patient data collected from the respective patient population. In the recent US Food and Drug Administration (FDA) Workshop on Seizure Detection, the agency instructed that, in “Indication for Use,” it has to include “In whom the device is intended to be used – specify intended population (age, patient group, seizure type).” Hopefully, advances in commercial seizure detection software over the next few years will improve performance and provide algorithms specific to different patient populations.
Related posts:
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
