Browse the corpus

Status epilepticus (SE) is a prolonged seizure in which the brain has exhausted the inhibitory mechanisms that result in the natural cessation of the majority of seizures. The most recent definition and classification of SE by the International League Against Epilepsy (ILAE) describes it as “a condition resulting either from the failure of the mechanisms responsible for seizure termination or from the initiation of mechanisms which lead to abnormally prolonged seizures (after time point t1).” It is further defined as “a condition that can have long-term consequences (after time point t2), including neuronal death, neuronal injury, and alteration of neuronal networks…” []. The new definition and classification highlight the relationship between seizure type and duration on the development of adverse sequelae. Four prospective population-based studies have been conducted that included both convulsive and nonconvulsive SE and used a definition incorporating a 30-min minimum duration. These report a minimum incidence of SE of approximately 10 to 15/100,000 in Switzerland [], Germany [], and Italy [], and 60/100,000 in the United States (Richmond, VA) []. Although incidence has increased over time, mortality rates have largely remained stable [, ]. A study of first episodes of SE showed that over half of cases occur in the absence of known epilepsy [].

Refractory SE (RSE) is defined as SE that persists after an appropriately selected and dosed benzodiazepine and second antiseizure drug [, ], although most studies examining RSE do not strictly follow the “appropriately selected” and “appropriately dosed” elements of this definition. In the only prospective study of RSE at the time of this writing, approximately 23% of patients admitted to the hospital with SE failed to respond to first-line and second-line therapy, wherein first-line therapy was a benzodiazepine and second-line therapy was either phenytoin, valproate, or levetiracetam []. Retrospective studies have reported rates of RSE between 31 and 43% [–]. “Super-refractory” SE (SRSE) is defined as continuous or recurrent seizures lasting 24 h or more following initiation of anesthetic medications for treatment of SE, including cases in which seizure control is initially achieved but is lost upon weaning the anesthetic drugs []. The best estimate of SRSE incidence is from Finland, where, in a study of 15 hospitals providing coverage for a population of 3.9 million, the annual incidence of SRSE was found to be 0.7/100,000 (95% confidence interval [CI] 0.6–0.9) []. Between 20 and 41% of people with RSE will progress to become super-refractory [, ].

Mortality rates in the literature have varied greatly based on inclusion and exclusion criteria—specifically, age, seizure classification, and etiology. Mortality rates range between 2 and 40% and are lowest when the etiology is fever, intoxication, drug/alcohol withdrawal, trauma, or subtherapeutic antiseizure drug levels from medication withdrawal/noncompliance [, , –]. RSE has a higher mortality compared with nonrefractory SE (~ 20% vs ~ 10%), and the mortality of SRSE is even higher (up to 40%) [–, ] []. Concerningly, mortality rates may be higher when evaluated six months to one year after discharge. In the Finnish population-based study referenced previously, hospital mortality was 7.4% (95% CI 0–16.9%), and 1-year mortality was 25.4% (95% CI 21.2–29.8%) [].

The cause of death in patients hospitalized with SE varies significantly and may be influenced by physician practice and patient population. Withdrawal of life-sustaining therapy (WLST) is the ultimate cause of death in many of these patients [], and it is likely that in these cases, WLST is preceded by prognostication of poor outcome by a clinician. Thus, it is of critical importance that prognostication be performed accurately on the basis of appropriately validated predictors. Validation of predictors of outcome in patients with SE is made challenging by the variable presentations and complications resulting from the seizures, systemic sequelae, and treatments. Prognostication during counseling is both essential and inevitable, however, and occurs routinely in intensive care units worldwide. The objective of these guidelines from the Neurocritical Care Society (NCS) and Deutsche Gesellschaft für Neurointensivmedizin (DGNI) is to outline broad principles of neuroprognostication in adults who are critically ill with SE, ensure that such prognostication is performed on the basis of the most reliable predictors available, and highlight opportunities for future research.

The scope of these Grading of Recommendations Assessment, Development and Evaluation (GRADE) guidelines is the prognostication of neurological outcome in adult patients who are critically ill with SE who receive guideline-concordant, standard-of-care treatment [, ]. Notably, these guidelines assume the use of standard-of-care treatment and should not be used to determine the value or intensity of therapeutic measures. The purpose of these guidelines is to provide evidence-based recommendations on the reliability of predictors of neurological outcome in adult patients who are critically ill with SE who have received standard-of-care treatment to aid clinicians in formulating a prognosis. The target audience consists of clinicians responsible for such counseling.

These guidelines provide recommendations on the reliability of select demographic and clinical variables as well as prediction models when counseling families and surrogates of patients with SE who receive standard-of-care treatment. To maintain consistency with published NCS treatment guidelines [], the panel considered the standard of care to include the use of continuous electroencephalographic (EEG) monitoring when available. Because continuous EEG is resource intensive and may not be widely available, serial intermittent EEG studies were considered an acceptable alternative when continuous monitoring was unavailable. We categorized predictors as reliable, moderately reliable, or not reliable. We based this categorization on a GRADE-based assessment of certainty in the body of evidence as well as effect size (quantification of predictor accuracy) across published studies, as shown in Table .

A key distinction exists between a reliable predictor of outcome in the context of counseling surrogates of patients requiring life-sustaining therapy and an independent predictor of outcome. An independent predictor fulfills one criterion—a statistically significant association with the outcome of interest in an appropriately conducted multivariate analysis. Independent predictors of outcome may be used for risk stratification, for selection of patients for targeted treatment (such as chemotherapy regimens for cancer), or as building blocks of clinical prediction models. A reliable predictor in the context of counseling the surrogates of patients requiring life-sustaining therapy must be independent but also meet a higher threshold, as described in the “” section. Confidence in the accuracy of the predictor should be sufficiently high to overcome concerns about the undesirable consequence of inappropriate WLST.

Reliable predictors, for the purposes of these guidelines, may be used to formulate a prognosis when the appropriate clinical context is present in the absence of potential confounders. These are predictors with a low rate of error in prediction of poor outcomes, with at least moderate certainty in the body of evidence. When the prognosis is formulated on the basis of one or more reliable predictors, the clinician may describe the outcome as “very likely” during counseling. Given the inherent limitations in neuroprognostication research, the clinician must nevertheless acknowledge the presence of uncertainty—albeit, low—in the prognosis during counseling. Moderately reliable predictors may be used for prognostication only when additional reliable or moderately reliable predictors are present in addition to the appropriate clinical context. These are predictors with a low rate of error in prediction of poor outcomes but with lower certainty in the body of evidence, frequently as a result of smaller studies that result in imprecision. When the prognosis is formulated on the basis of multiple moderately reliable predictors, the clinician may describe the outcome as “likely” during counseling but must acknowledge “substantial” uncertainty in the prognosis. Predictors deemed not reliable should not be used in prognostication. Variables deemed not reliable may be a component of reliable or moderately reliable prediction models.

Candidate predictors were selected based on clinical relevance and the presence of an appropriate body of literature. Candidate predictors and prediction models were considered “clinically relevant” if, in the subjective opinion of the content experts and guideline chairs, the predictor or components of the prediction models were (1) accessible to clinicians, although universal availability was not required, and (2) likely to be considered by clinicians when counseling patients and surrogates regarding prognosis. Predictors thought particularly likely to be considered by clinicians during prognostication were prioritized. An appropriate body of literature was considered to be present for clinical prediction models with at least 1 external validation study of at least 50 patients in addition to the initial report on development of the model (also with a minimum of 50 patients). Based on these criteria, the following candidate predictors were selected:

Etiology of SE: acute symptomatic or potentially fatal Classification of SE according to seizure type/semiology Treatment refractoriness: RSE defined as SE that persists after an appropriately selected and dosed benzodiazepine and second anticonvulsant drug Treatment refractoriness: SRSE defined as continuous or recurrent seizures lasting 24 h or more following initiation of anesthetic medications for treatment of SE, including cases in which seizure control is initially achieved but is lost upon weaning the anesthetic drugs Notably, we did not select specific patterns on EEG other than the persistence (refractoriness) of seizures as prognostic indicators in this systematic review, because they were no longer thought to be widely used by clinicians for this purpose and lacked a sufficient body of evidence. Specific EEG patterns have prognostic value in individual disease states, such as in survivors of cardiac arrest []. Studies have suggested that periodic discharges may, in particular, be associated with either a late stage of SE [] or severity of underlying injury [, ]. However, the prognostic value of these patterns across disease states that result in SE independent from etiology and duration of SE could not be meaningfully evaluated in this systematic review.

Status Epilepticus Severity Score (STESS) Epidemiology-based Mortality score in Status Epilepticus (EMSE) The population/intervention/comparator/outcome/time frame/setting (PICOTS) question was then framed for the specific candidate predictors as follows: “When counseling family members/surrogates of patients with SE, should [predictor, with time of assessment if appropriate] be considered a reliable predictor of [outcome, with time frame of assessment]?”.

The outcomes rated “critical” using the GRADE 1–9 scale were mortality (average rating 9) assessed at or beyond 3 months from hospital discharge, in-hospital mortality (average rating 9) assessed at hospital discharge, functional outcome (average rating 9) assessed at or beyond 3 months from hospital discharge, cognitive outcome (average rating 9) assessed at or beyond 3 months from hospital discharge, and psychiatric or behavioral outcomes (average rating 9) assessed at or beyond 3 months from hospital discharge. The most common time points for assessment of both mortality and functional outcome in the literature were hospital discharge, one month and three months after hospital discharge, last follow-up, and intensive care unit discharge. Functional outcome at hospital discharge was judged by the panel to provide a poor representation of meaningful long-term outcome because progressive improvement after discharge is well documented following SE [, ]. No studies that included cognitive, psychiatric, or behavioral outcomes met other full-text screening criteria for the systematic review. The panel was unable to provide recommendations on predictors of functional outcome because there was an insufficient body of evidence meeting our eligibility criteria that assessed this outcome at an appropriate time point. In addition, there was marked inconsistency in the evaluation of functional outcome, which was variably assessed in the literature using modified Rankin Scale (mRS)/Glasgow Outcome Scale cutoffs, comparison of functional status before and after SE based on before versus after mRS/Glasgow Outcome Scale, or poorly characterized criteria, such as “new impairments,””new disabilities,” or “functional decline after SE.” The panel could only provide recommendations on the reliability of predictors of in-hospital mortality, the only outcome with a sufficient body of evidence that met our eligibility criteria. The panel, however, recognized that death following SE frequently occurs as a result of WLST [] and that self-fulfilling prophecy was an important source of confounding in the body of evidence that addressed this outcome.

An in-depth description of systematic review methodology overall for the NCS-DGNI neuroprognostication guidelines project is in Supplementary Appendix , including the librarian search string used for this systematic review. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram is in Fig. . The initial librarian search was performed February 20, 2019, and encompassed the period of 1946 to the search date. An updated search was performed in March 2022. Full-text screening was performed with the following exclusion criteria: sample size less than 50, studies focused on a highly selected subgroup (such as immune-mediated SE), studies of predictors not established as independent with multivariate analysis, studies focused on a genetic polymorphism as a predictor, and studies of clinical prediction models that did not report model discrimination. Studies of neuroimaging and laboratory biomarkers were not included because of lack of an appropriate body of evidence at the time of full-text screening and lack of widespread use in clinical practice, respectively.

Studies with no restrictions on WLST and likely incorporation of predictors under investigation into clinical neuroprognostication during the course of the study were considered to have a high risk of bias from self-fulfilling prophecy. Studies from countries with restrictions or cultural limitations on WLST were judged to have a lower risk of bias from self-fulfilling prophecy []. A summary of individual studies of predictors is in Supplementary Appendix . The GRADE evidence profile and summary of findings (for in-hospital mortality are in Table .

For the purposes of these guidelines, predictors described as “reliable” have both a higher overall certainty in the evidence and greater effect size than “moderately reliable” predictors (Table ). For “reliable” predictors, one downgrade was permitted for risk of bias, but none were permitted for inconsistency, imprecision, or indirectness, and the overall quality of evidence had to be high or moderate. “Reliable” prediction models were required to demonstrate an area under the receiver operating curve (AUC) of > 0.8. Another important measure is model calibration, or the ability to correctly specify the probability of an outcome. Model calibration is typically reported as a goodness of fit, often using the Hosmer–Lemeshow test or with a calibration curve, slope, or intercept []. Assessment of calibration in at least one external validation study of adequate quality per our eligibility criteria was required for reliable prediction models. Single downgrades in each of the domains of risk of bias, imprecision, and indirectness were permitted for “moderately reliable” predictors, but a downgrade for inconsistency was not. In addition, “moderately reliable” prediction models were required to demonstrate an AUC > 0.7, and some miscalibration in some external populations was allowed. Predictors that did not fit “reliable” or “moderately reliable” criteria were classified as “not reliable.”

Accurate prediction of a poor outcome provides family/surrogates with information needed to make decisions aligned with the estimated wishes of the patient. In addition to providing family and surrogates with a greater degree of certainty in the anticipated course of events, it may provide a sense of closure and some degree of comfort from respecting the patient’s wishes. Inaccurate prediction of a poor outcome (i.e., a false-positive prediction of poor outcome) may lead to WLST in an individual who may otherwise have made a meaningful recovery. Because WLST almost always leads to death, the undesirable consequence of an inaccurate prediction of poor outcome was the primary concern during discussions on the reliability of a predictor.

The panel, including the patient representative, was in agreement that most patients and their surrogates would likely consider an inaccurate prediction of poor outcome that led to the death of a patient who might otherwise have had a reasonable recovery to be far more undesirable than a prolonged period of uncertainty in the outcome. Therefore, a high certainty in the evidence of predictor or prediction model accuracy was necessary to recommend consideration when counseling families and surrogates on prognosis in this context. However, the panel recognized that values and preferences vary greatly and that many consider prolonged dependence of life-sustaining measures, such as artificial enteral nutrition and an artificial airway extending over months or even years, to be unacceptable. This criterion was therefore challenging to apply to frame recommendations applicable to the broad range of patient and surrogate preferences in this disease.

The predictors evaluated in our systematic review did not require any expenditure of resources beyond the provision of standard medical care. Similar to other conditions, an accurate prediction of poor outcome in SE may avoid the extended use of resources, over days to years, in patients destined to suffer a poor outcome. The use of resources was therefore thought to favor consideration of a predictor or prediction model during prognostication when confidence in its predictive accuracy was high.

In accordance with recommendations of the GRADE network, these statements were considered by the panel to be actionable, supported by indirect evidence when appropriate, and essential to guide the practice of neuroprognostication []. The good clinical practice reflected in these statements lacked a meaningful body of direct evidence—typically because of insufficient clinical equipoise—but were considered by the panel to be unequivocally beneficial.

We recommend that clinicians establish with patient surrogates the most appropriate long-term goals of care in individual patients with SE. Such goals, which may range from control of seizures to a return to functional independence, should serve as the basis for prognostication in an individual patient (strong recommendation, evidence cannot be graded).

In the setting of SE, discussions with surrogates may be centered around or restricted to the prognosis for control of seizures, whereas the primary concern of patients and families may be long-term recovery. Goals of care vary widely, and clarity on acceptable outcomes is critical to prognostication. This must be established with surrogates through active listening and shared decision-making. In adults with the Lennox–Gastaut syndrome and other diseases that cause profound baseline neurological impairment, goals of care may be limited to the control of seizures and a return to home. Prognostication may therefore focus on the probability of termination of SE and the anticipated duration of endotracheal intubation or institutional care. In contrast, surrogates of a patient who is elderly and intubated with a subdural hematoma and SE may prioritize (based on a best estimate of the patient’s wishes) a long-term return to functional independence. The primary focus of prognostication in such a patient should therefore be on the likelihood of functional independence rather than termination of SE or absence of seizures alone.

Surrogates of patients with RSE/SRSE should be counseled that treatment measures such as prolonged sedative infusions and long-acting antiseizure medications (ASMs) may result in prolonged coma and dependence on life-sustaining measures, such as an artificial airway and artificial enteral nutrition, for a variable duration ranging from days to months. Surrogates should also be counseled that medical complications related to critical illness, hospital-acquired infections, and prolonged restriction of mobility are common in such patients (strong recommendation, evidence cannot be graded).

In patients with RSE/SRSE, it is critical that expectations for patient recovery be managed appropriately. Surrogates may expect rapid awakening following successful initial control of seizures, whereas neurological improvement may be considerably delayed. Recurrence of seizures on weaning of sedatives may necessitate prolonged infusion use and/or the initiation of ASMs with a prolonged duration of action (such as barbiturates). Patients with RSE and SRSE sometimes require prolonged use of life-sustaining therapy, including tracheostomy and percutaneous gastrostomy []. These patients are at risk for a variety of medical complications, such as hospital-acquired infections, venous thromboembolism, aspiration syndrome, and pressure-related injury [].

Surrogates of patients with RSE/SRSE should be counseled that a long-term recovery that reaches or approximates the patient’s baseline level of functioning is feasible. Factors such as underlying acute brain injury, comorbidities, nosocomial infections, and sequelae of critical illness may, however, limit functional recovery (strong recommendation, evidence cannot be graded).

Although patients with RSE/SRSE often require prolonged support, multiple studies indicate the possibility of recovery to long-term functional independence (or the patient’s functional baseline) following weeks or months of coma and life-sustaining treatment [, –]. Prediction of the duration of dependence on life-sustaining therapy and delayed awakening can be highly challenging. An appropriate period of treatment and observation for delayed recovery in such cases should be established based on a best estimate by surrogates of the patient’s willingness to undergo prolonged life-sustaining care.