The tragedy of over-read EEGs and wrong diagnoses of epilepsy

The misdiagnosis of epilepsy is common and has very serious consequences. It is in fact a life-changing diagnosis. A major contributor to the misdiagnosis of epilepsy is the tendency to over-read normal tracings as abnormal. In fact, the wrong diagnosis of seizures is sometimes based solely on the ‘abnormal’ EEG. Reasons for the common overinterpretation of normal EEGs are mostly related to the lack of standards or mandatory training in EEG, and the erroneous assumption that all neurologists are trained to read EEGs. Potential solutions include defining and ensuring EEG competency for neurologists who read EEGs and providing a confirmatory reading by an EEGer, as is done for ECGs.

The problem

Most neurology training programs include a neuropathology rotation, right? Would we as neurologists feel comfortable reading pathology slides and making tissue diagnoses (e.g., brain tumors) that determine prognosis and management? Why do we assume that any trained neurologist can competently read EEGs?

Cardiologists have it right. Any general practitioner (family medicine, emergency physician, pediatrician or internist) can read ECGs to make immediate clinical decisions. However, as a safeguard against serious errors, a confirmatory and ‘official’ reading by a cardiologist almost always follows. Even for neuroimaging studies (e.g., MRI) an official and confirmatory interpretation eventually comes from a (neuro)radiologist.

A good argument can be made that we should have the same for EEGs.

Typical month at an epilepsy center

• • BC is a 66-year-old woman who had a single episode of passing out upon standing, preceded by dizziness and diaphoresis. Her EEG was read as showing spikes and she was put on antiepileptic drugs (AEDs). The EEG in question was reviewed and was normal;

• • MF had a temporal lobectomy for intractable seizures, and was seizure free for over a year. She had no complaint, but had an EEG that was found to show spikes, so she was told she ‘still had seizures’. The EEG in question was reviewed and was normal, with innocent left temporal transients consistent with a breach and a photocell artifact;

• • AH is 70 years old and presented with several weeks of dizziness, headaches and falls. After her EEG showed ‘ongoing temporal lobe seizures’, she was treated with multiple AEDs. Symptoms worsened. The EEG in question was reviewed and was normal, with mildly sharp bitemporal activity;

• • LM is 58 years old and has episodes of left eye throbbing pain and numbness that can move to the right side, nausea, smell sensitivity, poor balance, whole body tingling and feeling hot and sweaty. Her EEG reportedly showed left epileptiform abnormalities, but her symptoms did not improve on AEDs. The EEG in question was reviewed and only showed mild left temporal slowing;

• • AA presented with episodes of dizziness and facial numbness lasting hours to days. Her EEG reportedly showed spikes and she was put on carbamazepine. Her symptoms did not improve. The EEG in question was reviewed and was normal;

• • MH is 6-year-old girl evaluated for staring and ‘crying’ spells. Her EEG reportedly showed bicentral sharp waves, but her episodes did not improve on two AEDs, so she was set up to receive a vagus nerve stimulator. After her episodes were shown to be nonepileptic, the EEG in question was reviewed and was normal, with prominent vertex waves of sleep;

• • AB was referred for intractable ‘seizures’ and found by EEG-video to have definite psychogenic nonepileptic attacks. A prior EEG had been read as a showing ‘seizures’, and when reviewed was completely normal, with only benign wicket spikes.

How serious is this problem?

The above case series, while it is not an elegant double-blind randomized placebo-controlled study, reflects what is happening daily to patients. It is difficult to quantify the problem, but all referral epilepsy centers see such patients, and frequently. Furthermore, for one patient like this whose diagnosis is eventually rectified, there are likely dozens who never are. Approximately 30% of patients seen at epilepsy centers for refractory seizures do not have seizures and have been misdiagnosed [1]. As illustrated here, many of them have histories not in the least suggestive of seizures, and have their diagnosis based largely (and sometimes solely) on an ‘abnormal’ EEG [2–8]. This goes against what is taught at every medical school and during neurology training (‘we do not treat the EEG’). It is also why many feel that ‘‘routine interictal EEG recording is one of the most abused investigations in clinical medicine and is unquestionably responsible for great human suffering’’ [5]. Yes, EEG can be bad for you.

The consequences of being misdiagnosed with epilepsy are obvious and serious [9]. When the diagnosis is based largely on an abnormal EEG, no amount of subsequent normal EEGs will ‘cancel’ the previous abnormal one, and the wrong diagnosis is very difficult to undo. Unlike a CT or MRI scan that is misread, simply repeating it and calling it normal will not cancel the abnormal one, because the EEG is subject to time sampling. Only rereview of the same ‘abnormal’ sample can cancel the misdiagnosis and undo the harm.

Reasons for the overinterpretation of EEGs

The most fundamental reasons are lack of training and inexperience (i.e., not seeing enough normal tracings and the range of normal variations) and not applying strict criteria to make sharply contoured waveforms epileptiform. The less experience, the lower the threshold for ‘abnormality’. This is a normal phenomenon. For the same reason, a neuroradiologist will have a higher threshold than general radiologists for calling an MRI abnormal, and a cardiologist will have a higher ECG abnormality threshold than an internist. Unfortunately, there is currently no Accreditation Council for Graduate Medical Education (ACGME) requirement for EEG, and one can become a full-fledged and certified neurologist with very little or no experience in EEG. This results in a low and ‘nervous’ threshold for abnormality, which leads to over-reading of normal patterns.

The importance of the EEG is overemphasized, and it is interpreted out of clinical context. The diagnosis of seizures relies mainly on a good history, which requires skill and time. Unfortunately, doctors and patients alike tend to have more faith in ‘tests’. It is easier (and more lucrative?) to order an EEG than to take a history. The combination of vague, nonspecific symptoms with an equivocal (weak) EEG abnormality is a common cause of inappropriate diagnoses of seizures [4–8,10].

One important human factor is that the reader is ‘trying too hard’ to find abnormalities because the patient had a ‘seizure’ (history bias). The frustration of having yet another normal EEG in a patient who had a seizure leads to the ‘looking too hard’ syndrome. Another factor may be that the technologist’s interpretation may bias the reading neurologist.

The EEG patterns that can be over-read include well-known normal variants, but in reality these are relatively rare and the vast majority of over-read patterns are either wicket rhythms or ‘nameless variants’, that is, simple fluctuations of sharply contoured background rhythms or fragmented α activity that do not fit into a well-described EEG waveform [6,8]. These ‘nameless’ fluctuations of normal backgrounds have been described under different names and are found on most EEGs [11–16]. Differentiating normal variants from meaningful spikes and sharp waves can at times be challenging, but helpful rules have been described [6–8,17–24]. Unfortunately, this is only well known by EEG–epilepsy specialists, and most EEGs are read by general neurologists (just like most ECGs are initially read by noncardiologists).

From a more technical point of view, one common error is the widespread overemphasis on ‘phase reversals’ and the common misconception that phase reversals are somehow indicative of abnormalities. Phase reversals do not indicate epileptogenicity or even abnormality [6–8].

Possible solutions & recommendations

Clearly many neurologists who read EEGs are not adequately trained to do so. The reality is that most EEGs ordered in routine clinical practice (typically for encephalopathy) have little or no impact on diagnosis, management and outcome. For the diagnosis of seizures and epilepsy, however, the consequences of misreading are enormous. So, how do we ensure that neurologists who interpret EEGs are qualified to do so? Based on what is routinely practiced in other specialties, there are three options (not exclusive).

First, use a backup or confirmatory interpretation. This is nearly universal for ECGs. ECGs are routinely read by noncardiologists (internists, emergency physicians, pediatricians) who can use their interpretations for real-time decisions in managing patients. However, there is almost always a backup system of confirmatory readings by a cardiologist. There is no such thing for EEGs, and by default all neurologists are assumed to be competent EEGers. Competencies are currently being defined for newer procedures (e.g., transcranial Doppler) to ensure that neurologists who interpret them are qualified, but EEG is assumed to be an easy and standard test, probably because it is not a new test. It is not. If all neurologists can read EEGs to manage their patients, there should be some safeguard with a confirmatory reading by an EEGer/epileptologist.

Second, if we are to assume that all neurologists who want to read EEGs are qualified to do so, there should be better, more and mandatory EEG training during neurology residency. It would be inconceivable to become an internist without demonstrating some proficiency in ECG interpretation, and in fact, this is specifically addressed in the cardiology and internal medicine fields [25–32]. Not so for EEGs. As is done for electrocardiography, the ACGME should require minimum EEG training in neurology residency, in terms of quantity (so many months, so many studies) and quality (under the supervision of EEGers/epileptologists). The ACGME currently has very few specific requirements, and ‘clinical neurophysiology’ is mentioned together with neuropathology and neuroimaging. However, there is an enormous difference because pathology specimens and neuroimaging studies are not read only by the neurologist. (We are, of course, welcome to look, but an official report comes from a trained pathologist or radiologist.) Clinical neurophysiology studies (EEGs, EMGs) are interpreted only by us neurologists. We have the final say.

Finally, we must define special competency for neurologists who choose to interpret EEGs. Our colleagues in cardiology are ahead of us in this regard. Such training requirements for noncardiologists who read ECGs in their practice are put forth and subjected to ongoing discussions [25–32].

More immediate and less politically charged measures could include the following:

• • ‘Conservative’ reading should be strongly emphasized during EEG training. All epileptologists agree that over-reading is more harmful than under-reading; in fact, this may be the only thing on which all epileptologists agree! ‘‘Hardly anyone with epilepsy will come to any harm from a delay in diagnosis, whereas a false-positive diagnosis is gravely damaging’’ [5]. ‘‘Much more damage is done by overdiagnosis than by underdiagnosis of epilepsy’’ [20];

• • To avoid the ‘looking too hard’ syndrome, EEGs should be classified blindly to (unbiased by) the history: this is in fact practiced and recommended by many EEG and epilepsy authorities, and should be incorporated into EEG guidelines. (The clinical interpretation should, of course, incorporate the history and the clinical setting, but the EEG itself can be classified without the history);

• • There is a need for authoritative guidelines on EEG interpretation. Unfortunately, the only existing American Academy of Neurology guidelines related to EEG are on esoteric or clinically irrelevant topics, such as quantitative EEG (1997 and 2003), and the EEG in the evaluation of headaches (1995). The American College of Cardiology works together with the American Society of Internal Medicine, the American College of Physicians and the American Academy of Emergency Medicine to raise standards for ECG [25–32]. Our professional organizations (American Academy of Neurology, American Epilepsy Society, American Clinical Neurophysiology Society) should address this politically difficult situation and work together to raise standards in EEG, in the interest of our patients.

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.