An Evolving Understanding of Interictal Epileptiform Discharges

Commentary

Electroencephalogram (EEG) is a key test toward establishing diagnoses and treatment plans for patients with seizures. Early electrical recordings from animal brains occurred in the late 1800s early 1900s, and the modern field of EEG was founded by Hans Berger after performing the first human EEG in 1924. Interictal discharges (IEDs) were first characterized in the 1930s, and the first clinical EEG laboratories arose in the United States by the 1940s.

Since then, great strides have been made toward refining and implementing EEG technology. In particular, evaluation of interictal IEDs represents one core task of EEG. Over the decades, considerable progress has been made regarding understanding the sensitivity and implications of IEDs.

For example, understanding the utility or sensitivity of identifying IEDs required numerous steps over time toward establishing a common definition. Early attempts (1966) from the Terminology Committee of the International Federation for Electroencephalography and Neurophysiology defined IEDs as waves “distinguished from background activity” (spike: ≤ 1/12th of a second; sharp: 0.2-0.5 s). ¹ Appropriately, this was soon called into question, given waveforms frequently fulfill such guidelines despite suspected limited clinical significance. Thus, attention was drawn to features resembling our current understanding, such as having an electrical field, surface negativity, and after going slow waves. ² Nomenclature was further codified over time, for example recent 2017 guidelines suggesting that an IED demonstrate at least 4 out of 6 criteria and not resemble known benign variants (eg, vertex waves, lambda, etc). ³ This has been approximately supported by subsequent investigation suggesting that approximately 4 to 5 out of 6 criteria may provide an optimal balance of sensitivity and specificity when assessed against Epilepsy Monitoring Unit ictal recordings. ⁴

Improving terminology allowed investigators to become better positioned to understand the clinical importance of IEDs. The National Institute for Neurological Disorders and Stroke has been instrumental in funding EEG-based work including the landmark 1970 study by Marsan and Zivin that described the prevalence of IEDs among patients with epilepsy. ⁵ Prior studies had often looked at the question from the standpoint of “any” EEG abnormality, including interictal epileptiform discharges or generalized or focal slowing. Whereas this study isolated epileptiform findings, which is likely critical due to the special importance of epileptiform findings in terms of predicting a patient's future seizure risk. The study identified 308 patients with “an unquestionable epileptic nature of their seizure disorder” based on history and observed seizures (excluding postoperative EEGs or patients with status epilepticus), who had at least three EEGs over time each, for a total of 1824 EEGs. They found that about 30% of patients had an IED on every one of their EEGs, 18% had no IEDs on any EEG despite repeated studies and despite their fairly certain clinical diagnosis of epilepsy, and the remaining 52% had some positive and some negative EEGs. Among the 27 patients followed for at least 1 year after an initial negative record, about 80% had a subsequently positive record. Several trends emerged within subgroups. For example, older age at EEG or older age at first seizure both predicted decreased chance of a positive finding (age 10-19: 32 of 96 = 60% mostly positive EEGs; age 40+: 20 of 51 = 39%). Additionally, 98% of patients with suspected temporal lobe epilepsy eventually had a positive EEG at some point, much higher than extratemporal epilepsy, though admittedly the study did not precisely describe how they determined localization. IEDs also dropped off with increasing durations since the last seizure. The ratio of patients with all positive to all negative EEGs was 5.8 when performed on the day of a seizure, whereas this ratio was only 0.39 when performed more than 30 days after the last seizure. Having all positive EEGs was similarly more common for patients with a higher seizure frequency (every day to few weeks) compared to a lower seizure frequency (less than once per year). Other features such as family history, semiology (“grand mal”), etiology, and nonparoxysmal background findings demonstrated less clear stories.

What can we take from this early work by Marsan and Zivin? First, this type of work has emboldened patients and clinicians to “try, try again.” We still routinely counsel patients and write in our EEG interpretations that a negative EEG does not exclude epilepsy. They found that many patients with an initially negative EEG can thereafter turn positive on subsequent studies. This was particularly the case in suspected temporal lobe epilepsy (per the authors’ words, epilepsy “below the sylvian fissure”) or older adults, and not clearly reduced by treatment decisions. This encourages that it still often appropriate to render a tentative diagnosis of epilepsy or start antiseizure medications despite a negative initial routine when the history remains compelling. Seen from another vantage point, a patient with repeatedly negative EEGs despite clinically suspected temporal lobe epilepsy could be seen with skepticism, or at least this would be an indication in the modern era to proceed with advantage complementary tests such as magnetoencephalography or positron emission tomography.

Clearly Though, This Work has Limitations

1. Confounding is a major limitation. Patients on medications or with a reason to obtain an EEG on the day of a seizure likely present more severe cases than those not yet treated or not felt to need a stat EEG.

2. EEG readers were not blinded to clinical history and thus could have been biased by such knowledge.

3. The “gold standard” in this case was clinical diagnosis. The issue is not only that we use a somewhat more inclusive definition today of epilepsy than they used in 1970, ⁶ but also that it seems fairly clear now that even expert clinicians are imperfect at distinguishing epileptic from nonepileptic episodes.

4. These data regarding the chance of IEDs in patients with a fairly clear epilepsy diagnosis are not particularly useful in many circumstances. Oftentimes, early in a patient's course, EEGs are obtained in a patient with intermediate pretest probability for having epilepsy or developing future unprovoked seizures. Thus, these data do not elucidate positive or negative predictive values or likelihood ratios, as would be most helpful to the clinician hoping to use EEG for risk stratification and treatment decisions, and patients without seizures may rarely have IEDs.

5. The study did not examine other probably important features affecting yield, such as sleep and recording duration.

6. Variability in whether EEGs are positive between studies could represent either true expressions that IEDs are intermittent thus may not be captured on every study, versus random or systematic noise in intrarater and interrater reliability. ⁷

7. While some distinction was made between seizure types, these general findings may not apply across the many epilepsies.

Subsequent research has shed further light on both the utility of routine EEGs and how to increase the likelihood of observing IEDs. Regarding risk stratification after a first seizure of life, IEDs are quoted as conferring a relative rate increase for seizure recurrence at 1 to 5 years of 2.2 as compared with patients without IEDs, ⁸ which can be further incorporated into multivariable prediction alongside the exam, semiology, and imaging. ⁹ Though not all EEGs are equal—while subsequent investigators have confirmed Marsan and Zivin's findings that repeated EEGs increase sensitivity, ¹⁰ yield also increases with the duration of recording. For example, studies evaluating latency to first IED comparing routine versus prolonged ambulatory monitoring have found that cumulative incidence climbs markedly within the first several hours of recording (eg, about 45% by 4 h to almost 90% by 24 h) but thereafter plateaus. ¹¹ Also routinely performed in EEG labs, sleep deprivation alongside activation maneuvers in the appropriate setting can also maximize yield. ¹²

We thank the National Institutes of Health for their long-term support for biomarker-based predictive research and continued development of EEG technology. The funding pipeline will remain critical to future innovation focused on transforming care and optimizing the accuracy and value of EEG.