A 24-year-old woman with new-onset seizure; a 50-year-old man with unwitnessed loss of consciousness; a 78-year-old man with altered mental status. What do these 3 patients have in common? All of them may be admitted to your hospital, and you might be wondering where electroencephalography (EEG) fits into the workup. For the hospitalist, knowing when to order the test—and when not to—poses a challenge. While a rudimentary understanding of electrocardiography (EKG) is required of every medical student in the United States, EEG is considered more esoteric. It's a qualitative test, with no “positive” or “negative” interpretation and findings that often seem vague or must be decoded by the ordering clinician. But hospitalists can learn to be that decoder.
EEG for beginners
Understanding the electrical basis of EEG is easier than many think because it mirrors a more familiar test, the EKG. Like EKG, EEG is a noninvasive tool that measures the electricity of the human body and displays it as a waveform. The x-axis represents time, and the y-axis represents amplitude. As EKG electrodes overlie cardiac structures (right and left ventricles), EEG electrodes overlie cerebral structures (F for frontal, T for temporal, P for parietal, O for occipital, and C for central sulcus). When you see high amplitude QRS complexes in the precordial leads on EKG (V1-6), you might predict left ventricular hypertrophy. Similarly, a high-amplitude spike wave in the temporal region on EEG suggests that a large pool of neurons may be firing synchronously. In the heart, synchrony promotes the heart's role as a highly efficient pump. The brain is not a pump, and in fact electrical charge fluctuates in a far less predictable manner, more akin to the “babbling” sound heard in a crowded room. However, there are some limits to the chaos: Background rhythms made by neurons fall within a “normal” range of 8 to 13 Hz when the patient is awake, while during sleep, slower rhythms and sleep-specific waveforms like spindles and vertex waves predominate. Diffuse slowing with the patient awake is the hallmark of encephalopathy, while transient abnormalities called interictal epileptiform discharges (including spikes, sharp waves, spike-and-slow wave complexes, and epileptiform activity) are the marker of a seizure tendency (a dysfunctional brain can produce both elements). These findings can occur randomly or in a more predictable (“periodic”) pattern. These epileptiform discharges can be thought of like premature cardiac beats: They may suggest pathology but must be evaluated in a clinical context. Up to 0.5% of the (asymptomatic) general population may have interictal epileptiform discharges (1). On an EEG, a seizure typically consists of evolving, rhythmic, epileptiform discharges over a period of time, most often with a clinical correlate.
Asymmetrical findings on EEG can also indicate pathology: a loss of amplitude (or attenuation) over just 1 hemisphere may indicate either injury to the cerebral cortex or an overlying fluid collection (e.g. subdural hematoma), while slow-ing isolated to 1 region or hemisphere may indicate structural abnormality or injury, which requires correlation with neuroimaging.
EEG interpretation is based on the balance of brain rhythms in varying states of alertness, with an eye towards symmetry, epileptiform discharges, and findings that would be expected given the age of the patient.
Think back to the 24-year-old woman with a new-onset seizure. Why order an EEG? Even in patients with known epilepsy, routine EEG is normal 50% of the time. It is nonetheless the appropriate screening test in patients with a first-time seizure. Despite its poor sensitivity, specificity of an EEG with interictal epileptiform discharges is 78% to 98% (2). Seizure is a clinical diagnosis; an EEG cannot tell a clinician whether a seizure has already occurred. Instead, the test is a prognostic tool to assess risk for seizure recurrence and diagnosis of epilepsy syndromes.
Before ordering an EEG for the first-time seizure patient, the clinician would typically exclude provoking factors, including derangements of electrolytes or glucose and use of or withdrawal from substances. A neurological exam, neuroimaging, and routine EEG are key in the workup for first unprovoked seizure. Along with a careful history (screening for family history and febrile seizures, among other factors), the results will determine if therapy with an antiepileptic drug is warranted. EEG is the best predictor of recurrence following a first unprovoked seizure. Recurrence rate ranged from 27% to 58% in 1 large study of patients with a first unprovoked seizure (3). In the 27% category were patients who subsequently had a normal EEG. In the 58% category were those whose EEGs demonstrated clear interictal discharges. In the middle (37%) were those with nonspecific abnormalities on EEG. Recurrence is also highest in the first 2 years following a first time seizure, and higher if the first seizure occurred at night rather than during daytime (4). Interictal epileptiform discharges can be focal—most commonly in the temporal lobe, in adult patients—or they can be generalized, involving the “brain as a whole.” Juvenile myoclonic epilepsy, contrary to its name, is the most common form of primary generalized epilepsy in adults. These patients typically have generalized tonic-clonic seizures without warning, as well as myoclonic jerks that often occur in the morning. Primary generalized epilepsies have characteristic EEG patterns (2.5-6Hz generalized spike-wave) that can be picked up on routine EEG; a diagnosis of a primary generalized epilepsy syndrome usually has treatment implications, since in adults, these are genetic conditions that do not go away on their own.
Loss of consciousness and syncope
In cases of loss of consciousness (such as the example 50-year-old patient), the role of EEG can be particularly tricky. The diagnosis of seizure is based primarily on clinical history, often by eyewitnesses. Even trained epileptologists will rarely see their own patients seizing, so differentiating a history of seizure from loss of consciousness may be difficult. Thus it is useful, before doing an EEG, to classify clinical history by pretest probability: low, medium, or high. A clinical episode considered highly probable for seizure may warrant treatment even without an EEG abnormality, if recurrent. On the other hand, an episode most suggestive of syncope (low pretest probability) would require a markedly abnormal EEG (generalized spike-wave discharges, for instance) to warrant medication treatment. And even then, clinicians must be aware that they are mainly treating the test and not the patient.
Syncope is caused by a sudden reduction in cerebral perfusion; it can affect up to 40% of the general population and is responsible for 3% of ED visits (5). Syncope can occur due to physiologic, autonomic, or cardiac causes. Seizure types that mimic syncope are less common in the adult population and when seen often have other associated focal features. When the clinical history suggests clear syncope, EEG is not useful given its low diagnostic yield. A study of 1,003 routine EEGs in patients with syncope yielded normal results in 90% of cases, nonspecific findings in 8%, and interictal epileptiform discharges in just under 2% (6).
However, in patients with recurrent syncope with no clear etiology, a negative cardiac workup, or atypical features, EEG might provide some diagnostic benefit. Expected EEG findings during syncope include diffuse slowing or attenuation, fitting with a momentary loss of brain perfusion. These findings can be correlated with cardiac telemetry, usually available with continuous EEG monitoring (cEEG), which can be performed either inpatient or outpatient. Recurrent loss of consciousness without a change in the EEG or EKG would be suspicious for psychogenic nonepileptic seizures, although sleep disorders like cataplexy can have similar features.
Altered mental status
Think now of the 78-year-old man with altered mental status. Encephalopathy is a clinical diagnosis associated with global cerebral dysfunction, and EEG is not part of the initial workup. However, in patients with persistently altered mental status despite maximal medical therapy and correction of underlying toxic-metabolic derangements, EEG becomes the test of choice, along with neuroimaging.
EEG allows clinicians to correlate the degree of slowing (mild, moderate, severe) with the depth of encephalopathy. In a patient who is unresponsive, a normal test suggests a psychiatric etiology. Also, nonconvulsive seizures and status epilepticus cannot be clinically differentiated from encephalopathy without EEG. Nonconvulsive seizures and status epilepticus can be associated with mortality rates as high as 51% (7) and are probably under-recognized in critically ill patients, with estimates between 8% and 20% (8). The high prevalence is probably multifactorial and can be related to sepsis, hypoperfusion, and hypoxemia leading to mechanical ventilation; circulating toxins both endogenous and iatrogenic; altered metabolism; and the wide use of certain antibiotics that can lower the seizure threshold (among them fluoroquinolones and beta-lactams, including carbapenem-based antibiotics). A remote neurological injury—stroke, for example—in the context of a new toxic-metabolic derangement can become a nidus for seizures. The Neurocritical Care Society published guidelines in 2012 that support the use of cEEG in the following scenarios: patients with recent seizures who have not returned to baseline, patients with coma (including following cardiac arrest), patients with intracranial hemorrhage or other acute neurological injury, patients with altered mental status and suspected nonconvulsive seizures and status epilepticus, and inpatients with a markedly abnormal routine EEG (9). However, inpatient cEEG is a limited resource in most institutions, and there is wide gap between recommendations and clinical practice, given that this test relies on availability of equipment, on-call EEG technologists, and expert interpretation (10).
There are a few instances in which the EEG may help tailor or even confirm the differential diagnosis in the patient with altered mental status. For example, “triphasic waves” are commonly a feature of hepatic or uremic encephalopathy, but they are not pathognomonic for these disorders, as had been supposed in the past. In the ICU setting, cEEG in coma following cardiac arrest can help assess both prognosis and the presence of seizures, which may complicate recovery (11). In a patient with subacute, progressive encephalopathy over weeks to months, particularly with myoclonus, a characteristic EEG with “periodic sharp waves” every 1.5 to 2 seconds can alert the clinician to the possibility of prion diseases such as Creutzfeldt-Jacob disease. Periodic discharges over just 1 hemisphere, known as “PLEDS” or periodic lateralized epileptiform discharges, can be seen in herpes encephalitis but are not pathognomonic for this disease. EEG alone does not distinguish mild encephalopathy from dementia; however, it can exclude a focal finding that may warrant further workup. EEG in the evaluation of persistently altered mental status should be used in conjunction with clinical history and examination, laboratory tests, and neuroimaging.
The inpatient use of EEG has grown in recent years, in part due to increased availability and expectation for cEEG according to guidelines from neurology and neurocritical care organizations. The use of cEEG in coma after cardiac arrest and acute brain injury, among other scenarios, is likely to grow further as cheaper, portable, or wireless cEEG units become available at community hospitals and as reading and interpretation can be done remotely.
For the hospitalist, EEG should be considered the gold standard test in the evaluation of new-onset seizures and in the management of status epilepticus, both convulsive and nonconvulsive. Though not indicated for the initial workup of syncope or altered mental status (with some exceptions), its utility as a noninvasive clinical tool is essential when loss of consciousness is recurrent or when encephalopathy is persistent.