Background

The incidence of seizures in general intensive care units (ICUs) ranges from 3.3% to 34% [1]. Risk factors include common diagnoses such as brain tumor, head trauma, stroke, history of seizure, electrolyte abnormalities, hypoglycemia, infections, and drug overdose or withdrawal [2]. Thus, ICU management of antiepileptic drugs (AEDs) is routinely practiced by intensive care providers.

Selection of the most effective AED with the least adverse events, however, can often be challenging. About a third of patients with seizures fail on monotherapy, necessitating two or more AEDs [3]. Several factors commonly seen in the critical care setting, such as polypharmacy, unpredictable medication pharmacokinetics, and implementation of a variety of nonpharmacological interventions, may lead to drug-drug interactions, elevated risk for drug toxicity, and subtherapeutic drug serum levels; these are discussed in this review.

Pharmacodynamics and pharmacokinetics of AEDs

Pharmacodynamics

AEDs depress abnormal neuronal firing by various mechanisms of action including altering ion channel activity, enhancing gamma-aminobutyric acid (GABA)-mediated inhibition, or reducing glutamate-mediated excitation. While some AEDs have a single mechanism of action, others have multiple mechanisms of action and, in some, the exact mechanism of action is not yet known (Table 1).

Table 1 Mechanism of action of antiepileptic drugs [113]
Full size table

When using multiple AEDs, it is reasonable to select medications with different mechanisms of action.

Pharmacokinetics (absorption, distribution, metabolism, excretion)

Alterations in normal physiology and the physical properties of medications can affect the rate and extent of enteral absorption of medications in the ICU, leading to a need for parenteral administration. Deranged gastrointestinal absorption may be expected in circumstances of decreased blood flow, intestinal atrophy, dysmotility, and interactions with enteral nutrition. Pharmacokinetic parameters of selected AEDs, including their oral bioavailability, are summarized in Table 2.

Table 2 Bioavailability and pharmacokinetic data of antiepileptic drugs [113]
Full size table

Changes in serum pH as well as both respiratory and renal failure affect the ionized state of many drugs, affecting their penetration across lipophilic-based membranes such as the blood-brain barrier. Bioavailability of active drugs is also affected by alterations in volume of distribution for hydrophilic medications, while hypoalbuminemia increases the unbound (free) fraction of highly albumin-bound medications.

Although drugs are commonly metabolized to less active compounds, prodrugs such as oxcarbazepine and fosphenytoin must be metabolized into their active forms carbamazepine and phenytoin, respectively. Drug metabolism generally occurs in two phases. Phase I involves nonsynthetic reactions to form a modified group. A cytochrome P450 (CYP450) enzyme is frequently involved in such oxidative reactions. Phase II involves synthetic reactions to conjugate the metabolite with an endogenous substance. Metabolism for most anticonvulsants occurs primarily in the liver and is dependent on hepatic blood flow, enzyme activity, and protein binding. Although critical illness often inhibits the CYP450 isoenzymes, drug metabolism may be enhanced over several days as in the cases of pentobarbital and phenytoin, resulting in potential subtherapeutic concentrations [4,5,6]. Induced hypothermia will also reduce the systemic clearance of many medications mediated by CYP450 (such as phenytoin or propofol), between 7 and 22% for every degree of Celsius below 37 [7].

Regardless of the route of administration, renal elimination of parent drugs or metabolites is the primary excretory pathway for most drugs. For patients on renal replacement therapy (RRT), the type of dialysis that is performed and the frequency/duration should also be considered (see section 8, special therapeutic considerations).

Dosing and therapeutic drug monitoring in the ICU setting

Dosing of AEDs should be individualized to achieve seizure control with minimal adverse effects. The 1-h postloading dose is commonly recommended as a time to measure peak serum concentration. Pharmacokinetic alterations are frequently observed in critically ill patients; hence, frequent therapeutic drug monitoring (TDM) may be required. Basic dosing and TDM recommendations for AEDs commonly used in the ICU are summarized below and in Table 3.

Table 3 Dosing and monitoring of commonly used antiepileptic drugs [8, 11, 15, 16]
Full size table

Phenytoin/fosphenytoin

Phenytoin has an average half-life of 24 h, but this ranges from 7 to 40 h, increasing with dose escalation due to its nonlinear kinetics [8]. Phenytoin is insoluble in water and is dissolved in a basic solution including ethylene glycol, the mixture being linked to tissue necrosis (“purple hand syndrome”) if extravasated [9]. Fosphenytoin, the prodrug of phenytoin, is water soluble and hence free of the toxic emulsion. This difference in solubility allows intramuscular and faster intravenous administration of fosphenytoin, but the need for plasma conversion to the active drug (phenytoin) results in a comparable time to peak plasma levels when compared with phenytoin administration itself [10].

Because phenytoin follows nonlinear or saturable metabolism pharmacokinetics, it is possible to attain excessive concentrations much easier than medications that follow linear pharmacokinetics. Phenytoin TDM is therefore clinically important in the critically ill and should be followed closely.

At normal serum levels, patients may experience minor central nervous system depression and adverse effects such as nystagmus, drowsiness, or fatigue. Beyond the normal target range, ataxia, slurred speech, and incoordination often occurs. Drug-induced seizure activity has been reported at concentrations over 50–60 μg/mL [8].

Typically, protein binding accounts for 90% of total plasma concentrations, hence the therapeutic range for unbound phenytoin (free) concentrations is 1–2 μg/mL. In patients suspected of having altered drug plasma protein binding, monitoring of free phenytoin serum concentration is of value.

Valproic acid

Although the accepted therapeutic range for total valproic acid concentration for seizure therapeutics is 50–100 μg/mL, levels up to 175 μg/mL have been suggested in cases of refractory status epilepticus (SE) cases. Concentration-related side effects include ataxia, lethargy, tremor, and coma [11]. The common adverse effects—thrombocytopenia and hyperammonemia (via carnitine depletion)—can often be limited by dose reduction and by carnitine replacement in the latter condition. Due to significant interpatient differences in valproic acid metabolism, there is a poor correlation between valproic acid dose and total serum concentrations [11, 12].

Valproic acid is highly (90–95%) protein bound and is saturable within the therapeutic range which results in higher unbound fractions at higher concentrations. Although not often monitored, a therapeutic free valproic acid range of 2.5–10 μg/mL can be used as an initial guide [11].

Phenobarbital

Phenobarbital has a long half-life of approximately 100 h, which limits its use when short-term AED use is desired. As with other AEDs, treatment of SE in particular may require higher than normal dosing, and this may reach upwards of even 10 mg/kg/day (serum levels of > 100 μg/mL) with solid efficacy demonstrated [13]. Concentration-related adverse effects of phenobarbital are sedation, confusion, and lethargy, with high doses leading to obtundation and respiratory depression [14]. As phenobarbital is only about 50% protein-bound, free drug monitoring is not warranted. However, in severe hepatic impairment (Child-Pugh score > 8) a decrease of 25–50% in the initial daily maintenance dose may be required [15].

Pentobarbital

Pentobarbital is often used in the ICU setting to treat SE or elevated intracranial pressure. Initiation of pentobarbital involves sequential bolus doses followed by a continuous infusion [16, 17]. The average half-life of pentobarbital in adults is reported to be about 22 h and it is 20–45% protein-bound. Serum pentobarbital TDM (reference range is 1–5 μg/mL) is of limited utility in determining treatment clinical response or toxicity; the direct measure of intracranial pressure (ICP) control or inducement of “burst suppression” by electroencephalography (EEG) are monitored instead. Serum concentrations, however, may be useful in assessing residual effects of pentobarbital-induced coma once discontinued [18].

Drug-drug interactions

AED-AED drug interactions

Administration of multiple AEDs is common in the ICU to manage refractory seizures and SE. Satisfactory seizure control is not achieved in 30–40% of patients with monotherapy, making it necessary to use two or more AEDs [19, 20]. Although synergistic anticonvulsant activity between medications may be desirable, concurrent use of multiple AEDs generally increases the risk of drug interactions. Table 4 lists common AED-AED drug-drug interactions.

Table 4 AED-AED drug interactions [113–116]
Full size table

AED-nonAED interactions

Interactions of AEDs with other agents is seen more frequently with drugs that are metabolized in the liver, and therefore are more common with older AEDs. Tables 5 and 6 describe AED-nonAED and nonAED-AED interactions.

Table 5 Interaction of selected antiepileptic drugs with commonly used medications in the intensive care unit [113, 115, 117–124]
Full size table
Table 6 Interaction of selected therapeutic classes with antiepileptic drugs [115, 117, 127–132]
Full size table

Seizure prevention in the ICU

Traumatic brain injury

The incidence of post-traumatic seizures accounts for 6% of overall symptomatic epilepsy [21,22,23] and is estimated to occur in up to 15% and 53% in civilian and military populations, respectively [24, 25]. Clinical studies have divided post-traumatic seizures into either immediate (< 24 h), early (within 1 week, with a seizure incidence of 2.1–16.9%), or late (with an incidence of 1.9% to > 30.0%) [22, 23, 26]).

Early seizures occur in about 10–15% of patients not treated with anticonvulsants, and this incidence can be reduced to about 2–3% with the use of phenytoin [27]. There is, however, little evidence that such use reduces the occurrence of late seizures or the incidence of death and neurologic disability [28]. It thus appears that anticonvulsants are suppressive during the early period following traumatic brain injury (TBI), but are not prophylactic for the time period beyond the first week. The American Academy of Neurology (AAN) [29] and the Brain Trauma Foundation [30] advocate prophylactic use of AED only during the first 7 days after a head injury [31, 32]. Phenytoin has been shown to decrease the incidence of early post-traumatic seizures [30] and to be more cost-effective than levetiracetam [33]. Although levetiracetam was shown in small early studies to be as effective as phenytoin [33, 34], there are recent data suggesting a lack of any prophylactic action in a comparative study of levetiracetam versus no prophylaxis [35]. A trial that incorporated a blinded design and the use of continuous electroencephalography (cEEG) monitoring demonstrated no difference in early or late seizure incidence between drugs, but did favor levetiracetam over phenytoin in the 6-month Glasgow Outcome score measure [36].

The use of valproic acid, as compared with phenytoin, failed to demonstrate benefit and its use correlated with higher mortality [37]. However, a recent Cochrane analysis failed to discriminate between phenytoin and other anticonvulsants with regards to reduced incidence or mortality [37].

Ischemic stroke

Seizures following ischemic stroke occur within an incidence range of 4% to 14% [38,39,40]. According to a large recent registry from France, early seizures are not associated with higher mortality or unfavorable outcomes at 1 month or at 1 year [41]. Patients with late-onset seizures have a greater than twice the risk for subsequent stroke, although the causal relationship is unclear [42].

Given the rather low incidence of seizures following most ischemic stroke syndromes, neither the American Heart Association (AHA) nor the American Stroke Association guidelines recommend prophylactic use of anticonvulsants following ischemic stroke [43]. However, some patients may be at greater risk for seizures. An increase in seizure risk has been reported following cardioembolism [39], hyperglycemia [41], intravenous thrombolysis with alteplase [44], in patient with SE following poststroke seizures, and either large cortical or hemorrhagic infarction [38, 45,46,47,48].

If “prophylaxis” is considered, then the agent used should be selected for its adverse effect profile. In recent years there has been some interest in ascribing statins as a “first-line” anticonvulsant in stroke patients, although supportive data are by association only and this type of intervention is not readily practiced [49, 50].

Intracerebral hemorrhage

In contrast to ischemic stroke, intracerebral hemorrhage (ICH) portends a somewhat higher risk of seizures, ranging from two- to seven-times that from ischemic stroke [38, 47, 51, 52]. If cEEG is performed during the acute admission period, nonconvulsive epileptiform activity is commonly observed and may affect more than 20% of ICH patients [53,54,55,56]. As expected, lobar or supratentorial ICHs have the highest incidence of seizures; those in the posterior fossa have almost zero [38, 54]. Surgical evacuation confers an increased incidence, and subarachnoid or subdural blood also increases the risk for seizures [57].

Prophylactic anticonvulsant treatment in ICH patients is currently not strongly advocated. cEEG is suggested by some for ICH patients with a depressed mental status out of proportion to that expected for the severity of the stroke [58]. Levetiracetam has become the most widely prescribed agent despite a lack of evidence [59], likely due to its ease of use. Uncontrolled study data have suggested that phenytoin confers worse outcomes than others (primarily against levetiracetam) [60,61,62], but these data are not supported by other recent studies [63, 64].

Subarachnoid hemorrhage

Early and late-onset seizures are observed in patients with subarachnoid hemorrhage (SAH). Risk factors include the severity of SAH, concomitant intracerebral or intraventricular blood, and treatment with craniotomy [65]. A recent study noted a 10% incidence of seizures, with almost half (47%) occurring within the initial 24 h [66]. Interestingly, the use of prophylactic anticonvulsants did not reduce the overall risk. As seizures dramatically increase cerebral blood flow and blood pressure, rebleeding from a recently ruptured aneurysm during seizures may pose a serious risk.

Current guidelines from the AHA state that the use of anticonvulsants for a brief period is reasonable, without specific drug recommendation [67]. Beyond the immediate posthemorrhagic period, routine use of anticonvulsants is advocated by the AHA only in patients with a high risk for delayed seizures, such as prior seizure, intracerebral hematoma, intractable hypertension, infarction, or middle cerebral artery aneurysm. The Neurocritical Care Society (NCS) specifically recommended against the routine use of phenytoin for prophylaxis [68].

Brain tumors

The incidence of seizures in cerebral malignancy broadly ranges from about 10% to as high as 90%, and is correlated with tumor type (glioma and oligodendroglioma being amongst the highest risk) and also the rate of progression, with slow growing tumors tending towards higher seizure incidence [69]. Prophylactic use of AEDs in brain tumors remains controversial. There are strikingly little data in support of seizure prophylaxis for brain tumor patients [70,71,72], and guidelines by the AAN and the European Federation of Neurosciences do not support this practice [73, 74]. Since these guidelines were written prior to the widespread use of cEEG and prior to the introduction of newer, less toxic anticonvulsants (levetiracetam, lamotrigine, lacosamide, and so forth), some have suggested that such a recommendation should be reassessed [75].

Whenever patients require adjuvant chemotherapy, it is prudent to avoid the use of potent CYP3A4 coenzyme inducers such as carbamazepine, phenytoin, and phenobarbital because of the risks of compromising concurrent chemotherapy [74]. Based on more recent evidence, more patients are being managed with levetiracetam than phenytoin or sodium valproate [71], but evidence of superior efficacy remains absent [72].

Cerebral venous and sinus thrombosis

Seizures manifest as a common presentation, occurring in 29% to 50% of patients with cerebral venous sinus thrombosis [76, 77]. Approximately a third to one half of seizures occur early—as the presenting symptom or within 2 weeks of diagnosis, [77,78,79]. Despite the lack of controlled studies, prophylaxis appears prudent where forthcoming seizures appear highly probable. There is some evidence that early seizures are predictive of a remote epilepsy condition [80] but, in and of themselves, seizures in the context of cerebral venous sinus thrombosis do not appear to be associated with death or 6-month worse outcomes [81].

Seizure control in the ICU

The widespread use of cEEG has made it apparent that seizures are more common in the ICU patients than previously thought, with estimates for seizures and SE ranging from 19% to 27% [82, 83]. Bauer and Trinka introduced the distinction between nonconvulsive status epilepticus (NCSE) proper and comatose NCSE [84]. This distinction is both important and useful as it acknowledges that treatment as well as prognosis are dependent on the underlying epileptic syndrome in the former, yet on the underlying cause of the coma in the latter. In the ICU patient population, the electrographic seizure patterns can be less distinct and harder to identify. Thus, a body of literature on the nature of, and the best treatment approach for, the “ictal-interictal continuum” has emerged and now dominates the literature. For recent reviews, see Sivaraju and Gilmore 2016 [85] and Rodríguez et al. 2016 [86].

As a general principle, the current consensus is that treatment of even a single ICU seizure should occur in order to prevent escalation into SE. Once airway, breathing, circulation, and “dextrose” (the “ABCDs”) have been addressed, pharmacologic seizure treatment should occur immediately. Several authors suggest an algorithm that escalates from first-line to second-line treatment in the time span of 30 min [87, 88]. Once urgent first- and second-line agents are prescribed during the first 30 min, persistent SE should be considered refractory, and this should prompt the use of intravenous anesthetic therapy. SE is a neurological emergency and swift intervention is of the essence. While convulsive and nonconvulsive SE should be initially approached in the same way, once the need for intubation arises it is generally accepted that NCSE may warrant a less aggressive and more individualized approach to treatment with anesthetic agents (third-line agents) compared with convulsive SE.

Available evidence uniformly suggests that benzodiazepines remain the drugs of choice for the immediate control of seizures of any kind. The efficacy of intravenous lorazepam was demonstrated by Treiman et al. in 1998 [89]. Lorazepam was most effective in the initial treatment of convulsive SE when compared with phenobarbital and diazepam plus phenytoin, although the latter were also effective. Other routes for the administration of benzodiazepines have successfully been explored, particularly for the therapy of seizures encountered outside the ICU. Silbergleit et al. reported that intramuscular midazolam was at least as safe and effective as intravenous lorazepam for prehospital seizure cessation [90], while McIntyre et al. demonstrated the efficacy of buccal midazolam [91].

Although an open-label use study purported to suggest equipoise between lorazepam and levetiracetam in the treatment of SE [92], a randomized, double-blind, placebo-controlled trial (SAMU Keppra Trial) failed to show benefit from adding intravenous levetiracetam to clonazepam in the treatment of generalized convulsive SE [93].

To monitor for ongoing seizure activity after convulsions may have ceased, cEEG is crucial. Second-line therapy for SE should be initiated within 30 min if first-line treatment has failed. Agents used as second-line therapy include fosphenytoin, valproic acid, phenobarbital, and levetiracetam. Evidence for the superiority of one over the other is generally lacking [94]. Nonetheless, phenytoin has been the most studied drug in SE and is therefore often listed as the preferred choice over the others. Valproic acid is notable for being the only agent that has shown a trend towards superior efficacy in some studies but this has not been confirmed in a large-scale trial [95]. Levetiracetam has a favorable side-effect profile and is therefore increasingly prescribed as a second-line drug. Phenobarbital is often avoided due to its long half-life, although data support its efficacy [89]. Lacosamide has recently shown promise in small case studies [96], while brivaracetam is currently under investigation [97].

If second-line agents are ineffective, treatment is typically escalated to anesthetic agents. For patients in NCSE, however, the decision whether or not to intubate and sedate the patient will usually be made on a case-by-case basis. There is some evidence to suggest that uncontrolled NCSE will result in brain damage over time [98]. Furthermore, DeLorenzo et al. noted that persistent NCSE after convulsive status may carry a worse prognosis than other forms of nonconvulsive status and perhaps should be treated more aggressively, although the presence of NCSE may simply be a marker for more severe brain injury [99]. Agents used for continuous infusion are midazolam, propofol, and pentobarbital, with no data suggesting the superiority of one over the other [100]. Ultimately, it is important not only to treat the patient with an anesthetic agent but to also continue maintenance therapy with appropriate AEDs to allow transition to a well-tolerated, long-term antiepileptic regimen.

Drugs lowering seizure threshold

Polypharmacy is common in the ICU and can contribute to adverse drug effects such as seizures by lowering the seizure threshold or via drug-drug interactions that may reduce the blood concentration of an AED the patient may already be receiving (see the next section). Antibiotics, psychotropic agents, and analgesics have been all associated with seizures [101, 102] (Table 7).

Table 7 Selected list of medications associated with lower seizure threshold [113]
Full size table

Special therapeutic considerations

Extracorporeal removal of antiepileptic drugs

Antiepileptic drug removal by renal replacement therapies

The renal route of elimination, low protein binding, low volume of distribution, and molecular weight are physiochemical factors that impact the propensity for drug removal via dialysis (Table 8). Unfortunately, kinetic data on optimal dosing of AEDs in RRT are not readily available. Clearly such patients benefit from careful TDM and a multidisciplinary approach. There are extensive expert reviews available in the medical literature [103, 104].

Table 8 Factors impacting drug removal by renal replacement therapies
Full size table

Intermittent hemodialysis

The removal of most AEDs by intermittent hemodialysis (IHD) is usually well characterized in the drug’s labeled prescribing information or other references. Two common drugs that are removed by IHD and for which data recommendations exist are levetiracetam and lacosamide. In addition, while phenobarbital is predominantly hepatically metabolized, some reports suggest partial elimination by IHD prompting several experts to suggest supplemental dosing after IHD and monitoring of serum drug levels [103, 105, 106].

Continuous renal replacement therapy

There is insufficient evidence to recommend clear dosing guidelines for AEDs in continuous renal replacement therapy (CRRT), and more research is needed to appropriately guide therapy. AEDs that are eliminated via IHD will also be eliminated by CRRT, often times even more so (depending on the duration of CRRT as well as the mode and flow rates) thereby requiring higher doses. For drugs where serum levels are available, TDM is indicated. Estimations of drug removal can be performed to estimate the removal of medications by CRRT based on flow rates, as well as drug and filter properties. The reader is referred to other resources for a more comprehensive discussion on these methods [107, 108]. A recent review of the literature proposed dosing recommendations based on AED protein binding and the extent of renal versus hepatic elimination [104]. This is a reasonable approach, although clinicians should also factor in CRRT flow rates, the clinical condition of the patient, and other factors into their clinical decision making accordingly.

Antiepileptic drug removal by plasmapheresis

In the neurocritical care unit, plasmapheresis (PLEX) is often employed to manage neuroautoimmune diseases. Unlike RRT, which removes the free fraction of a drug, PLEX removes whole plasma which includes both the free fraction of a drug as well as the protein-bound portion. Thus, a key factor in determining the whole body removal of a drug with PLEX is its volume of distribution. Other factors can include the duration/frequency of PLEX and exchange volume, as well as the rate of intercompartmental equilibration. In general, drugs with low volumes of distribution (< 0.2 L/kg) predominantly reside in the vascular compartment and PLEX would be expected to remove a significant portion of the total body stores of these drugs. As for AEDs with high volumes of distribution such as phenytoin, PLEX appears to remove only about 2.5–10% of total body phenytoin [109, 110].

Antiepileptic drug removal by extracorporeal membrane oxygenation

Extracorporeal membrane oxygenation (ECMO) can impact the pharmacokinetics of drugs in a number of ways. ECMO circuits are well known to sequester drugs given the large surface area of the tubing and membranes which can result in an increase in the volume of distribution initially [111], while later releasing the drug into the circulation resulting in an unpredictable effect [112]. Lipophilic and highly protein-bound drugs are considered to be particularly vulnerable, but other physiochemical factors such as molecular weight, protein binding, and ionization may all play a role [113]. In addition, ECMO is typically associated with reduced drug clearance as a result of alterations of end-organ perfusion. ECMO is also often combined with some form of RRT, further complicating predictability. One case report suggests ECMO has little impact on the disposition of levetiracetam (which has a low volume of distribution and low protein binding) [114]. Where available, clinicians should monitor serum levels to guide dosing in patients on AEDs.

Conclusion

Selection of the most appropriate AED in the ICU setting can be challenging for a variety of reasons. Older AEDs such as phenytoin, valproic acid, and phenobarbital are often used by clinicians due to their familiarity with these agents, intravenous formulations, and availability of evidence in certain clinical scenarios. Despite this, adverse effects of these agents, drug-drug interactions, and the need for TDM may limit their use. Newer agents such as levetiracetam and lacosamide are gaining popularity due to their relatively safe AED profile, fewer drug-drug interactions, and lack of need for TDM. The efficacy of these agents for seizure prophylaxis and as second-line treatment for SE, however, should be further evaluated in large randomized clinical trials.