PHILADELPHIA—Avoiding an inappropriate diagnosis of epilepsy is of critical importance, and it is well accepted among epileptologists that more damage is done by overreading than by underreading EEGs. Nevertheless, epileptologists frequently encounter EEGs that have been overread, resulting in erroneous diagnoses of epilepsy. “But the interesting thing here is neurologists who don’t read EEGs all the time, rather than erring on the side of underreading, err on the side of overreading,” said Selim R. Benbadis, MD, at the 2001 Annual Meeting of the American Epilepsy Society. “That’s more damaging. You make diagnoses of epilepsy in people who have very benign abnormalities, and like other misdiagnoses, it’s very hard to undo.” Dr. Benbadis is Associate Professor of Neurology in the Departments of Neurology and Neurosurgery and Director of the Comprehensive Epilepsy Program and Clinical Neurophysiology Laboratory at the University of South Florida and Tampa General Hospital.

Dr. Benbadis and colleagues reviewed a series of misread EEG records to identify trends in overinterpretation. Cases were collected during a two-and-a-half-year period from adult patients who were diagnosed with psychogenic nonepileptic seizures (PNES) but who had carried a previous diagnosis of epilepsy. Diagnosis of PNES was made with EEG video recording. Patients with evidence of coexisting epilepsy on EEG video were excluded. A total of 127 patients were diagnosed with PNES. Of these, 41 (32%) had a history of epileptiform EEG, and “we were able to obtain 14 actual recordings for review. All were from general neurologists or non-epileptologist EEGers,” Dr. Benbadis noted. The tracings were then reviewed by a board-certified electroencephalographer.

Based on the report and record description, Dr. Benbadis and colleagues were able to precisely identify the waveform interpreted as abnormal. Overread epileptiform abnormalities found in the EEGs included: wicket spikes (one), hypnagogic hypersynchrony (one), and hyperventilation-induced slowing (one). In the other 12 records, overread patterns were simple fluctuations of sharply contoured background rhythms or fragmented alpha activity.

Well-defined normal variants are rarely the problem in EEG overreading, Dr. Benbadis observed. “There were only three of them, and there were no tracings with small sharp spikes, psychomotor variants, or 14 and six positive spikes. Physiologic or extraphysiologic artifacts were also not a problem. Most overinterpreted patterns are fluctuations of normal rhythms that do not have a name. Variants of alpha activity are by far the main offender, especially the widespread, unusually sharply contoured variants, and especially during early drowsiness, when alpha waves become fragmented and isolated,” he elaborated. “Benign variants are not that common that they make a big difference. These nameless variants are common. When we read EEGs or train in EEGs, there are certain tricky normal variants, such as ‘small sharp spikes’ and ‘wicket spikes,’ which we learn to recognize. The interesting thing here is that most of the overinterpreted patterns are not those. They are benign variations of the background. They’re clearly meaningless.” However, because much of EEG interpretation “is a matter of training and experience, these nameless variations of normal need to be emphasized during teaching,” he added.

The causes of EEG overinterpretation are unclear and probably complex. According to Dr. Benbadis, they may include trying “too hard [to find abnormalities] when you get the history and you suspect seizures. My feeling is that this happens a lot. One of the philosophies of reading EEGs that I teach is to read the EEGs blindly, without history. In other words, you should at least classify the record without knowing anything about the history.” Not applying strict criteria to sharp transient epileptiforms was another possible explanation.

“Based on the feedback I received, I can tell you that every epileptologist who does full-time epilepsy and reads EEGs most of the time faces this [overinterpretation] once a month if not once a week. It’s a very common problem,” Dr. Benbadis said. “Of 10 patients I see that tell me they’ve had an abnormal EEG, I might get one that has the record for me to review. Most of the time they don’t bring the record. That’s a severe limitation, and there are very few ways around it.”

Dr. Benbadis hopes to continue examining the problem of EEG overinterpretation. “My goal would be, if this catches attention, to study it prospectively. I want to increase the numbers, because the proportion we get is small. We could only obtain a few of the EEGs requested as abnormal, so this underestimates the magnitude of the problem. An estimation of frequency was not possible, but the problem of EEG overreading is almost certainly underreported. It definitely contributes to the erroneous diagnosis of epilepsy, and it also complicates the management of PNES because patients hold onto the previous ‘abnormal’ EEG as irrevocable proof of the diagnosis of epilepsy, which is thus more difficult to correct. I would like to take the next 100 patients we monitor and diagnose as nonepileptic, take the proportion of those who have had abnormal EEGs, and look at it prospectively,” he said. “It’s difficult to write about because it’s politically unpleasant, if not politically incorrect. But it needs to be addressed. It’s an important problem.”

NR