Electrographic seizures and electrographic SE are frequent and are often associated with poor outcome in critically ill patients [11]. In a study of 201 adults in a medical ICU without known acute neurologic injury, 21 patients (10%) had electrographic seizures , 34 (17%) had periodic epileptiform discharges, 10 (5%) had both, and 45 (22%) had either electrographic seizures or periodic epileptiform discharges [12]. Electrographic seizures were electrographic-only (not clinically detectable) in most patients (67%) [12]. After controlling for age, coma, renal failure, hepatic failure, and circulatory shock, the presence of electrographic seizures or periodic epileptiform discharges was associated with death or severe disability, with an odds ratio of 19.1 (95% CI, 6.3–74.6) [12]. In a large series of 5949 adult (>17 years) patients who underwent mechanical ventilation (as a proxy for ICU stay) and in whom a routine EEG or a cEEG was performed, the use of EEG was independently associated with lower in-hospital mortality (OR = 0.63, 95% CI, 0.51–0.76) with no significant difference in cost or length of stay [13].

However, electrographic seizures are not always an independent predictor of poor outcome. In a series of 247 adult (>17 years) patients presenting to the emergency department with seizures who either died in the emergency department or were admitted to the ICU, ten patients died in hospital, and nine were discharged to hospice for a total of 19 (7.7%) patients with a poor prognosis [25]. The causes of death (septic shock in three patients, cocaine-associated intracranial hemorrhage in two, cardiac death in two, ruptured cerebral aneurysm in one, acute ischemic stroke in one, and ethylene glycol ingestion in one) suggest that the primary etiology and not the seizures was largely responsible for mortality [25]. Independent risk factors for poor outcome were early intubation (OR 6.44, 95% CI, 1.88–26.6) and acute intracranial disease (OR 5.78, 95% CI, 1.97–18.6) [25]. In contrast, presence of SE was not associated with a poor outcome in this series [25].

Whether electrographic seizures worsen outcome independent of the underlying etiology or whether they are just a biomarker of a more severe underlying lesion is a matter of ongoing discussions and study, and there are significant hurdles to appropriately answer this question best. First, the frequency of electrographic seizures reported in the literature is inherently biased due to confounding by indication as most patients only undergo cEEG when clinically indicated. A survey of 330 neurologists showed that the most common indications for performing cEEG monitoring in critically ill patients were altered mental status with or without seizures , subtle eye movements, acute brain lesion, and paralyzed patient in the ICU [26], but indications vary between and within centers. Second, patients might be misclassified as not having electrographic-only seizures because they occur prior to initiation of monitoring or after cEEG is discontinued. cEEG is most commonly used for at least 24 h if there are no electrographic seizures and, when detected, maintained for at least 24 h after the last seizure [26]. However, in certain populations such as patients with non-traumatic subarachnoid hemorrhage, the first seizure can occur later than 48 h [27] and might be missed when monitoring routinely for 24–48 h. Last, the impact of seizures on outcome might be etiology specific; electrographic seizures might cause marked outcome worsening in subarachnoid hemorrhage but not in brain tumors. Studies that analyze heterogeneous and broad etiologic categories jointly might bury associations in individual subgroups. In the following sections, we summarize the impact of electrographic seizures on outcome in different subgroups of critically ill patients (Tables 2.1 and 2.2).

EEG background abnormalities and electrographic seizures are common in patients with sepsis [28]. In a series of 71 adults with septic shock, 43 patients underwent EEG for clinical signs of coma, delirium, or seizure, and 13 (30.2%) had electrographic seizures [29]. Based on the limited literature available, sepsis is a risk factor for electrographic seizures , and when patients present with seizures and sepsis, outcomes are poor. In a series of 201 patients admitted to the ICU for acute neurologic injury, sepsis on admission was the only independent predictor of electrographic seizures , and electrographic seizures were independent predictors of death or severe disability at hospital discharge with an OR of 19.1 (95% CI, 6.3–74.6) [12]. In another series, 100 of 154 adults in a surgical ICU developed sepsis, and among these patients 24 (15.6%) had electrographic seizures and 8 of 24 (33.3%) patients had electrographic SE [30]. Electrographic seizures (including electrographic SE) were independently associated with poor outcome (death, vegetative state, or severe disability) at hospital discharge, with an OR of 10.4 (95% CI, 1–53.7) [30]. Further data are needed to evaluate the impact of electrographic seizures during sepsis on long-term outcome and to elucidate whether treatment modifies outcome.

Electrographic seizures are frequent following traumatic brain injury (TBI) and are often associated with poor outcomes . In a series of 94 adults with moderate to severe TBI, electrographic seizures occurred in 21 (22.3%) patients including 6 with electrographic SE [31]. In this study, 11 of 21 (52.4%) patients had electrographic-only seizures [31]. Patients with electrographic seizures had similar outcomes as compared to patients without electrographic seizures , but the six patients with electrographic SE died (100% mortality), compared to 24% mortality in the non-seizure group [31], suggesting that the presence of SE (and not only the presence of seizures ) was associated with worse outcomes .

Electrographic seizures following TBI are also frequent in children, especially during the first 2 years of life [32, 33]. In a series of 144 children with TBI and cEEG, 43 (29.9%) patients had electrographic seizures , 17 of whom had electrographic-only seizures [33]. The presence or absence of electrographic seizures or status epilepticus did not correlate with discharge outcome, classified in only three broad categories: death, discharge to rehabilitation facility, and discharge to home [33]. In a series of 87 children with TBI and cEEG admitted to the ICU, 37 (42.5%) patients had electrographic seizures , 14 of whom had electrographic-only seizures [32]. Outcomes at hospital discharge—measured more precisely with the King’s Outcome Scale for Childhood Head Injury—were worse in patients with clinical and subclinical SE, electrographic-only seizures , and electrographic SE [32].

Worse outcomes might simply reflect a more severe underlying etiology. However, some studies suggest that electrographic seizures contribute to poor outcomes . In a series of 140 patients with moderate to severe TBI and cEEG, 32 (22.9%) patients had electrographic seizures [34]. A subgroup of patients had volumetric MRI at baseline (within 2 weeks of TBI) and 6 months after TBI [34]. Among these, six had electrographic seizures and were compared with a control group of ten patients matched by Glasgow Coma Scale, CT lesion, and occurrence of surgery [34]. Patients with electrographic seizures had greater hippocampal atrophy on follow-up as compared to those without seizures (21% vs 12%, p = 0.017) [34]. Further, hippocampi ipsilateral to the electrographic seizure focus demonstrated a greater degree of atrophy as compared with contralateral hippocampi (28% vs 13%, p = 0.007) [34]. These data suggest that post-TBI electrographic seizures cause long-term anatomic damage [34].

The underlying mechanism by which post-TBI electrographic seizures cause worse outcomes may be related to the dysregulation between excitation and inhibition. In a mouse model of TBI, glutamate signaling was elevated and GABAergic interneurons were reduced following controlled cortical impact—an experimental equivalent to moderate to severe TBI [35]. In parallel, spontaneous excitatory postsynaptic current increased, and inhibitory postsynaptic current decreased after controlled cortical impact [35]. Similar dysregulation is described in humans. In a series of 17 adults with severe TBI, a microdialysis probe measured extracellular glutamate during the first week post-TBI [36]. Transient elevations in extracellular glutamate occurred during periods of decreased cerebral perfusion pressures of less than 70 mmHg, but also during seizures with normal cerebral blood perfusion [36]. Dysregulation of brain metabolism may elevate intracranial pressure and may lead to worse outcomes . A study in adults with moderate to severe non-penetrating traumatic brain injury evaluated cEEG and cerebral microdialysis for 7 days after the traumatic brain injury [37]. Matched for age, CT lesion, and initial Glasgow Coma Scale, ten patients with electrographic-only seizures (seven with electrographic SE) were compared with ten patients without electrographic seizures [37]. Patients with post-traumatic electrographic-only seizures experienced a higher mean intracranial pressure, a greater percentage of time of elevated intracranial pressure, and a more prolonged elevation of intracranial pressure beyond post-injury hour 100 as compared with patients with no post-traumatic electrographic seizures [37]. Similarly, the lactate/pyruvate ratio was elevated for a longer period of time and more often in patients with post-traumatic electrographic-only seizures [37]. In ten patients with electrographic-only seizures , a within-subject design compared the time periods 12 h before seizure and 12 h after seizure onset within the same patients [37]. Seizures led to episodic increases in intracranial pressure, in lactate/pyruvate ratio, and in mean glutamate level [37]. These findings have been confirmed on a series of 34 patients with severe traumatic brain injury in whom brain microdialysis showed that the lactate/pyruvate ratio increased during seizures and pseudoperiodic discharges, but not during non-epileptic epochs [38]. These results suggest that intracranial pressure and lactate/pyruvate ratio go up in response to electrographic-only seizures , and not vice versa. Electrographic-only seizures may therefore not only represent a simple biomarker of brain damage, but may contribute to further brain damage, at least in the context of traumatic brain injury.

Prior studies suggest that seizures contribute to worse outcomes by causing additional damage. Therefore, prophylactic AED use to reduce seizure burden, and to improve outcomes , has been contemplated. However, a Cochrane review of randomized clinical trials found that AED prophylaxis in TBI reduced the risk of early seizures (within 1 week of TBI) but not late seizures or mortality [39]. In summary, post-TBI electrographic seizures lead to further brain damage, but to date there is lack of evidence that electrographic seizure control after TBI improves outcomes .

Electrographic seizures are particularly frequent and appear relatively late in patients with subarachnoid hemorrhage. Electrographic seizures occurred in 8 of 69 (11.6%) patients with non-traumatic high-grade subarachnoid hemorrhage [27]. In this study, the initial 17 patients underwent cEEG for clinical suspicion of electrographic seizures , which occurred in 3 (17.7%) patients [27]. The following 52 patients underwent cEEG as part of a protocol for subarachnoid hemorrhage, regardless of clinical suspicion, and electrographic seizures occurred in 5 (9.6%) patients [27]. Among the 35 patients monitored per protocol and without a clinical suspicion of seizures , electrographic seizures occurred in 3 (8.6%) patients [27]. While high clinical grade of the subarachnoid hemorrhage was associated with poor outcome, the presence of electrographic seizures was not [27]. In a series of 402 patients with subarachnoid hemorrhage, seizure burden was independently associated with outcome so that every hour of seizure on cEEG was associated with an OR of 1.1 (95% CI, 1.01–1.21) to 3-month disability and mortality [40].

In contrast, other studies showed an association between electrographic seizures and poor outcome. In a series of 479 adult patients with subarachnoid hemorrhage, 53 (11%) had electrographic seizures [10]. Patients with electrographic seizures had increased clinical and laboratory inflammatory biomarkers, and the degree of inflammation was an independent predictor of electrographic seizures [10]. On univariate analysis, both inflammatory burden during the first 4 days and the presence of electrographic seizures were associated with poor outcome [10]. But after correction for other potential confounders, only electrographic seizures remained a predictor of poor outcome [10]. Further, mediation analysis showed that the effect of inflammation on outcome was mediated through the presence of electrographic seizures [10]. In summary, this study suggests that blood products in the brain may trigger an inflammatory cascade which causes electrographic seizures and, eventually, poor outcomes [10].

The concept that inflammatory cascades cause or contribute to seizures , and these impact outcome, is clinically relevant because it implies that anti-inflammatory products may reduce seizures . In fact, targeted anti-inflammatory treatments have controlled seizures or brain damage in animal models. In a rat model of inflammation—induced either by intestinal inflammation which increases tumor necrosis factor alpha (TNFα) or by direct infusion of TNFα—neuronal excitability increased with severity of inflammation. Central antagonism of TNFα prevented increase in seizure susceptibility [41]. Further, in a rat model of subarachnoid hemorrhage, the administration of IL-1RA—an interleukin-1 (IL-1) antagonist—reduced blood-brain barrier breakdown and the extent of brain injury [42].

Despite promising animal models, anti-inflammatory therapy for seizure prevention in subarachnoid hemorrhage is not ready for clinical use. A more realistic goal in subarachnoid hemorrhage may be outcome prediction based on cEEG results. In a series of 116 patients on cEEG following subarachnoid hemorrhage and with 3 months functional follow-up (measured with the modified Rankin Scale), the absence of sleep architecture (OR, 4.3) and the presence of periodic lateralized discharges (OR, 18.8) independently predicted poor outcome [43]. Additionally, outcome was poor in all patients with lack of EEG reactivity or state changes within the first 24 h, generalized periodic epileptiform discharges, or bilateral independent periodic lateralized epileptiform discharges and in 92% of patients with nonconvulsive SE [43].

cEEG may not only predict, but it may also modify outcome in patients with subarachnoid hemorrhage by early detection of delayed cerebral ischemia. In a series of 34 patients with subarachnoid hemorrhage, 9 (26.5%) patients developed delayed cerebral ischemia [44]. Visual analysis of cEEG demonstrated new onset slowing or focal attenuation in seven of nine patients with delayed cerebral ischemia (77.8%) [44]. Further, decreases in alpha power to delta power ratio identified patients with delayed cerebral ischemia with a very high sensitivity and a reasonable specificity: a cutoff of six consecutive recordings with more than 10% decrease in alpha power to delta power ratio from baseline yielded a sensitivity of 100% and a specificity of 76%; a cutoff of any single measurement with more than 50% decrease in the ratio yielded a sensitivity of 89% and a specificity of 84% [44]. In summary, cEEG detects electrographic seizures and/or delayed cerebral ischemia in patients with subarachnoid hemorrhage and in the appropriate clinical setting may improve outcomes .

Seizures occur frequently after stroke, particularly when it involves a hemorrhagic component or conversion from ischemic to hemorrhagic. In a study of 6044 patients with stroke, 190 (3.1%) patients experienced clinical seizures within the first 24 h, and these were more frequent in hemorrhagic than in ischemic stroke [45]. On univariate analysis, patients with seizures had a higher 30-day mortality rate than patients without seizures (32.1% vs 13.3%) [45]. Clinical seizures might represent only “the tip of the iceberg” as electrographic-only seizures might be overlooked without cEEG monitoring. In a series of 109 patients with stroke, electrographic seizures within 72 h occurred in 21 (19.3%) patients: 18 of 63 (28.6%) patients with intracranial hemorrhage, and in 3 of 46 (6.5%) patients with ischemic stroke [46]. On univariate analysis, posthemorrhagic seizures worsened the NIH Stroke Scale and increased midline shift [46]. However, on multivariate analysis, seizures did not independently predict outcome [45, 46]. Similarly, clinical seizures within 30 days of non-traumatic supratentorial hemorrhage occurred in 57 of 761 (7.5%) patients, but seizures were not independent predictors of in-hospital mortality [47]. In concordance with these results, electrographic seizures occurred in 32 of 102 (31%) adults with non-traumatic intracerebral hemorrhage who underwent cEEG monitoring [48]. In this series, 20 (20%) patients died in the hospital, and 5 patients had poor neurological outcome including coma, persistent vegetative state, or minimally conscious state [48]. Independent factors associated with poor outcome (defined as death, vegetative or minimally conscious state) at hospital discharge were coma at the time of hospital admission (OR 9, 95% CI, 2.4–34.3), intracranial hemorrhage volume of 60 ml or more (OR 4.4, 95% CI, 1.2–15.7), presence of periodic epileptiform discharges (OR 7.6, 95% CI, 2.1–27.3), periodic lateralized epileptiform discharges (PLEDs) (OR 11.9, 95% CI, 2.9–49.2), and focal stimulus-induced rhythmic, periodic, or ictal discharges (SIRPIDs) [48]. In this series, lower systolic blood pressure on admission was protective (OR 0.9, 95% CI, 0.9–1 per mmHg), but the presence of electrographic seizures did not influence outcome [48]. In summary, after non-traumatic intracranial hemorrhage, clinical seizures occur in approximately 3–8% of patients, but many seizures are subclinical, and EEG detects electrographic seizures in approximately 20–30% of patients. However, based on available data, seizures were not independent predictors of outcome.

Further, seizures were also not clearly associated with acute deterioration following stroke. In a series of 266 patients with non-traumatic intracranial hemorrhage, early neurological deterioration occurred in 61 (22.9%) patients and was associated with an eightfold increase in poor outcome; however, seizures were not a predictor of early neurological deterioration [49].

Considering more detailed outcome features, late seizures are weakly associated with subsequent development of epilepsy . In a series of 1897 patients with stroke, seizures occurred in 168 (8.9%) patients: 28 of 265 (10.6%) patients with hemorrhagic stroke and 140 of 1632 (8.6%) patients with ischemic stroke [50]. Recurrent seizures occurred in 47 of 1897 (2.5%) patients with late first seizure onset as sole risk factor for epilepsy following ischemic stroke. In this series, the authors did not find risk factors following hemorrhagic stroke [50]. Mechanistically, early seizures may frequently reflect brain irritation by blood products, while late seizures may often be related to gliosis and scarring [51]. In a series of 110 patients who underwent clot evacuation after intracerebral hemorrhage , the frequency of seizures was particularly high, occurring in 41%: early-onset seizures (within 2 weeks) in 31% and late-onset seizures in 10% [52]. Independent predictors of early-onset seizures were volume of hemorrhage, presence of subarachnoid hemorrhage, and subdural hemorrhage, and independent predictors of late-onset seizures were subdural hemorrhage and increased admission international normalized ratio (INR) [52]. These results suggest that the severity of the bleeding is related to seizure development. Therefore, these authors recommended prophylactic antiepileptic therapy for patients with severe intracranial hemorrhage [52]. However, the value of AEDs in intracranial hemorrhage remains unclear. In a study where 5 of 295 (1.7%) patients with intracranial hemorrhage had seizures , the use of AEDs was an independent risk factor for poor outcome after 90 days [53]. The deleterious effects of stroke and SE appear to act synergistically, so that mortality is three times higher when there is SE on an ischemic stroke compared to when there is only an ischemic stroke [54]. In summary, poststroke seizures are not independent predictors of outcome [50, 51], and their treatment does not necessarily improve prognosis [53]. However, the presence of SE may worsen prognosis, especially in ischemic stroke [54].

Seizures after cardiac surgery are relatively frequent, especially in neonates, and are associated with worse short-term outcome in most studies. In a series of 2578 patients who underwent cardiac surgery, clinical seizures occurred in 31 (1.2%) cases [55]. Compared to patients without seizures , patients with seizures experienced an almost fivefold increase (29% vs 6%) in in-hospital mortality, a higher incidence of all major postoperative complications, and a lower one-year survival rate (53% vs 84%) [55]. In contrast, a study of 101 adult patients (>18 years) who underwent subhairline cEEG after cardiac surgery found that 3 (3%) had electrographic seizures —two electroclinical and one electrographic-only—but the presence of electrographic seizures did not affect morbidity or mortality [56]. Of note, lack of correlation between electrographic seizures and outcome in this particular study may have been related to a limited EEG montage utilizing a four-channel subhairline EEG applied in the frontal and temporal regions with a typical sensitivity of 68% (95% CI, 45–86%) and a specificity of 98% (95% CI, 89–100%) for detecting seizures as compared to a regular 10–20 electrode placement [56, 57]. Further, only three patients had seizures , limiting statistical analyses in the series [56].