How Confident Are We of the Diagnosis of Epilepsy?


  • This paper was presented in condensed form at the Annual Course of the American Epilepsy Society at Washington DC on the 4th of December 2005 by C. A. van Donselaar.

  • The Dutch Study Group of Epilepsy in Childhood (DSEC) consists besides the authors, of O. F. Brouwer (University Medical Centre Groningen), A. C. B. Peters and O. van Nieuwenhuizen (University Medical Centre Utrecht), E. Peeters (Medical Centre Haaglanden, the Hague), and A. T. Geerts (Erasmus Medical Centre, Rotterdam). The DSEC received several grants from the Dutch National Epilepsy Foundation.

  • Conflict of interests: none.

Address correspondence and reprint requests to Dr. C. A. van Donselaar, Medisch Centrum Rijnmond-Zuid, Olympiaweg 350, 3078 HT Rotterdam, The Netherlands. E-mail:


Summary:  The diagnosis of a first seizure or epilepsy may be subject to interobserver variation and inaccuracy with possibly far-reaching consequences for the patients involved. We reviewed the current literature.

 Studies on the interobserver variation of the diagnosis of a first seizure show that such a diagnosis is subject to considerable interobserver disagreement. Interpretation of the electroencephalogram (EEG) findings is also subject to interobserver disagreement and is influenced by the threshold of the reader to classify EEG findings as epileptiform. The accuracy of the diagnosis of epilepsy varies from a misdiagnosis rate of 5% in a prospective childhood epilepsy study in which the diagnosis was made by a panel of three experienced pediatric neurologists to at least 23% in a British population-based study, and may be even higher in everyday practice. The level of experience of the treating physician plays an important role. The EEG may be helpful but one should be reluctant to make a diagnosis of epilepsy mainly on the EEG findings without a reasonable clinical suspicion based on the history.

 Being aware of the possible interobserver variation and inaccuracy, adopting a systematic approach to the diagnostic process, and timely referral to specialized care may be helpful to prevent the misdiagnosis of epilepsy.

Diagnosing epilepsy is usually straightforward, but in everyday clinical practice often doubts arise. Not everyone who faints with myoclonic jerks has epilepsy and the differential diagnosis of “fits, faints, and funny turns” is extensive. The diagnosis of a first epileptic seizure or epilepsy has proven to be subject to error (1–3). The mainstay of the diagnosis is a good eyewitness account of the episode(s), but this information may be not available, be incomplete, misleading, or rather vague, and may not correspond to the definition of the various epileptic seizures. The results of the additional investigations like the electroencephalogram (EEG) may be noninformative, provide information that is difficult to interpret or may seem to be contradictory to the clinical data. The level of experience of the treating physician is also important (3,4).

For any clinical diagnosis two questions should be considered. The first is whether others agree with the diagnosis: the interobserver variability, interobserver variation, or reliability. The second question is whether the diagnosis is correct: the accuracy or validity. To prove the accuracy of a diagnosis, a gold standard is needed. However, for most patients with epilepsy seen in everyday care, no gold standard is available. Clearly it is not possible to admit all patients for EEG-video monitoring to assess that the episodes are really epileptic seizures. An excellent response to antiepileptic drug (AED) treatment does not preclude misdiagnosis.

The diagnosis of a first seizure or epilepsy is often subjective. Patients may be misdiagnosed as suffering from epilepsy or the diagnosis of epilepsy may be missed. The consequences of a misdiagnosis may be far-reaching: psychosocial and socioeconomic problems such as driving restrictions and loss of employment may occur, the episodes may continue, an effective treatment may be withheld or patients may be treated with ineffective drugs often in high dosages or with polytherapy. If a diagnosis of epilepsy is incorrectly made in a patient with cardiac arrhythmia, the consequences may be serious, even life-threatening (5).

This paper reviews the diagnostic process, the interobserver variation and the accuracy of the diagnosis of a first seizure or epilepsy. The paper reflects our opinion based upon the main papers on this subject and our own work in this field.


To make a diagnosis of epilepsy the patient should have had two unprovoked epileptic seizures. The definition of epilepsy is not completely clear in case of sporadic unprovoked seizures. For example, if a person has a single unprovoked seizure and then a recurrence 5 years later, the diagnosis of epilepsy may not be justified. If a patient with a congenital hemiplegia or previous stroke has one epileptic seizure, epilepsy will be diagnosed by many neurologists (6).

The first question is whether the episode was an epileptic seizure or a nonepileptic event. If it was an epileptic seizure, what type of seizure (7)? Additional investigations may determine the etiology (third step). The fourth step is using all information to classify the epilepsy syndrome (8). Then, one can fully advise the patient and decide about starting treatment and the selection of AEDs.

This paper focuses on the first step: was it an epileptic seizure? The main problem is that we do not have validated criteria to diagnose an epileptic seizure. We have a classification scheme to categorize the seizure (7). But which descriptive elements of the episode warrant the diagnosis of a seizure? What if the episode only mentions loss of consciousness with some myoclonic jerks and loss of urine? Syncope, for example, may be accompanied by myoclonic jerks in many patients, by other motor symptoms like head turning and oral movements or attempts to sit up, and by incontinence for urine (9). Cardiac arrhythmias such as occur in the prolonged QT-interval syndrome, may mimic epileptic seizures (5). Validated descriptive criteria to distinguish between these entities are lacking.

Additional investigations and especially EEG findings may be helpful, but may also be misleading. If the clinical description points to a certain seizure type or epilepsy syndrome and the EEG findings correspond, the EEG is a useful tool for the classification of the seizure type and the epilepsy syndrome. However, the EEG may be normal in a substantial proportion of patients with epilepsy and the EEG may show epileptiform abnormalities in about 1% of aircrews undergoing routine EEGs as part of the screening procedure (10). The prevalence of epileptiform discharges in normal children is 3.5% and even higher in children with other neurological or behavioral disorders (e.g., ADHD) (11,12). Those with epileptiform abnormalities have only a slightly increased risk to develop epilepsy (10). Misinterpreting normal EEG patterns as epileptiform abnormalities can lead to the misdiagnosis of epilepsy (13). The EEG may show abnormalities that seem to contradict the clinical diagnosis. For example, the clinical description may point to a partial onset whereas the EEG shows generalized epileptiform abnormalities. If the diagnosis is doubtful on clinical grounds, but the EEG shows epileptiform abnormalities, one may be tempted to make a (mis)diagnosis of epilepsy. Decision rules are lacking.

Interobserver reliability

If the interobserver variation of a diagnostic process or classification scheme is poor, such a system cannot be accurate. Hence, one should improve the interobserver variation first. Unfortunately, this phase is often skipped, probably explaining the poor performance of many of these diagnostic systems or classification schemes when applied by others or used in different circumstances.

Studies on the interobserver variation of the diagnosis of an epileptic seizure or epilepsy often focus rather artificially on one aspect such as different neurologists assessing written descriptions of the episodes, reading of EEGs, evaluating neuroimaging studies, or interpreting videos of possible epileptic seizures. The results may be influenced by the way the information is presented, the instructions to the observers (make only a diagnosis if you are absolutely sure), the views and the levels of experience of the observers. Poor reliability results in these studies may not automatically point to similar poor performance in real life since the real life situation allows consideration of all available information. An eyewitness account is the mainstay of diagnosing epilepsy. The interobserver agreement of history taking has not been investigated (14).

First seizure

The interobserver variation of the diagnosis of a first seizure has been studied in adults and in children (15,16). Written descriptions of the episodes were presented to (pediatric) neurologists and the interobserver agreement was assessed. The observers did not have the EEG results. Kappa statistics were used to correct the observed agreement due to chance. Interobserver agreement in adults was moderate (kappa 0.58) when the neurologists were asked to give their clinical opinion. It was substantial (kappa 0.73) if descriptive criteria were used. For example, unconsciousness without an apparent cause with incontinence of urine is not considered to be an epileptic seizure compared with unconsciousness without an apparent cause plus a tongue bite (15). In children, the interobserver agreement was poorer: according to clinical judgment it was moderate (kappa 0.41) and improved only slightly when comparable descriptive criteria were used (kappa 0.45) (16). So, the diagnosis of first seizure is subject to considerable interobserver disagreement.


The EEG can help classify episodes as epileptic seizures but only if the eyewitness account is very suspicious of epileptic seizures. EEG interpretation is subject to interobserver disagreement. Disagreement on whether or not the EEG shows epileptiform abnormalities is substantial and is influenced by the threshold of the reader for classifying the EEG as epileptiform. Making a syndrome diagnosis will probably also vary among different readers (17–19).


To actually prove that a diagnosis is correct a gold standard is needed. In 26% of the patients who were referred because of intractable epilepsy to a tertiary center, another diagnosis was made often based upon long-term EEG-video monitoring (20). Good clinical practice is to refer patients in case of intractability to assess whether the diagnosis of epilepsy or the classification of the epileptic syndrome is correct. Hence, these referrals cannot be considered to be a random sample of patients being cared for in nontertiary centers. Their referral to the tertiary center was correct and one may not assume beforehand that this 26% reflects the actual “misdiagnosis rate” in everyday clinical practice. Alarming however is a misdiagnosis rate of epilepsy of 23% found in a population-based study in the United Kingdom, whereas in another 12% of patients the diagnosis proved to be disputable (1). The authors assume that their results are representative for the entire United Kingdom. Major concern was also raised in the United Kingdom after reviewing the medical records of 214 children with epilepsy from the practice of one English pediatrician (21). Over one-third of children diagnosed as having epilepsy were thought not to have epilepsy, and below one-third was probably overtreated. Similar rates of misdiagnosis were found among generalist pediatricians with an interest in neurology (4,21).

Not all patients can be studied with long-term EEG-video monitoring. Several other strategies have been adopted in studies on the accuracy of the diagnosis of a seizure. One approach is “wait and see” to assess whether new information will cast doubt on the initial diagnosis. If seizures do not recur with or without AEDs (i.e., no new information), then it may be assumed that the initial diagnosis of epilepsy was correct. Yet, such an assumption may be false. Another approach is to have the medical records reevaluated by one or more experienced neurologists, but this approach is still not an objective substitute for a gold standard.

First seizure

The accuracy of the diagnosis of a first seizure in adulthood was good in a study of 165 adults with a first seizure (22). The diagnosis was made exclusively on the account of the episodes by a panel of three neurologists. All patients were followed according to the wait and see policy. In 6% doubts about the initial diagnosis arose mainly because a cardiac arrhythmia was found during follow-up.

The Dutch Study of Epilepsy in Childhood was a prospective cohort study of children referred with a possible first seizure or epilepsy (2,23). The diagnosis was made by three pediatric neurologists exclusively on the description of the episodes for the children with a single event. In most of the children with multiple events the diagnosis was based only on the description of the events, but in 11% the results of both the history and EEG were combined. In all children a standard EEG was recorded, followed by an EEG after partial sleep deprivation if the first EEG showed no epileptiform discharges. A diagnosis of a first seizure was made in 170 children of whom 94 had epileptiform EEG abnormalities (2). In none of these children doubts arose about the initial diagnosis during follow-up, but only 53 of them suffered a recurrence. In 54 children, the panel classified the episode as a single uncertain event including 14 children (21%) with epileptiform EEG discharges. In four children (7%) a diagnosis of epilepsy was made during a follow-up of 1 year. Three of these belonged to the group of 14 with epileptiform EEG abnormalities. Thus, the diagnosis of a first seizure was probably quite accurate in these two studies in which the diagnosis was made by a panel of experienced adult and pediatric neurologists.


A diagnosis of epilepsy was made by the panel in 412 children of whom 285 had epileptiform abnormalities in their EEG (2). In 19 (5%) children there were doubts about the initial diagnosis. The diagnoses finally made were uncertain episodes (6), pseudo-seizures (2), syncope (2), daydreaming (3) and acute symptomatic seizure, breath holding spell, hair dressers syncope, TIA (in a child with moya-moya syndrome), anxiety disorder, and alternating hemiplegia in one child each. The misdiagnosis rate was only 2% for the children in whom the EEG had shown epileptiform abnormalities and 11% in the group without epileptiform abnormalities. In 124 children, the panel decided that the nature of the events was uncertain despite epileptiform EEG abnormalities in 27 children. In seven children (6%) finally a diagnosis of epilepsy was made; one in the group with epileptiform abnormalities (4%) and six (6%) in the group without epileptiform abnormalities.

In two British series the initial diagnosis of epilepsy was reevaluated. Misdiagnosis of epilepsy was at least 23% in a population based study and 16% in a hospital based study (1,3). The misdiagnosis rate for neurologists (5.6%) was much lower than for nonspecialists (19.3%) in the hospital based study (3). In a series of 684 children referred for paroxysmal disorders, the events were classified initially as an isolated seizure [51], epilepsy [83], possible epilepsy [90], or nonepileptic events (243 children) (24). All cases were reviewed at 6–30 months after the initial diagnosis by the same physician. Of the 90 children with possible epilepsy 31 were reclassified as having epilepsy (34%), none of the children diagnosed as having epilepsy was reclassified. The remarkable accuracy, in contrast to other studies, may be explained by the relatively high number of events classified initially as uncertain or by the design of the study with the same physician making the initial and the final classification. The authors advocated the use of a diagnostic category of “uncertain events” or “unclassified paroxysmal events” instead of “possible epilepsy,” and to follow these children without a straightforward diagnosis to prevent an unnecessary misdiagnosis of epilepsy.


Excluding the diagnosis of epilepsy because of normal EEG findings or making a diagnosis mainly on the EEG findings may be an abuse of the EEG (25,26). In the Dutch study, children with a diagnosis other than epilepsy were excluded. If a diagnosis of a single seizure was made in all of the remaining 224 children referred with a single event exclusively on the basis of epileptiform abnormalities on the EEG, 11 of the 108 children (9%) would have been misdiagnosed (2). If epilepsy was diagnosed in all children with multiple events only on the basis of EEG epileptiform abnormalities, the error rate would have been 10% (31 of 312 children) (2). The EEG appears helpful to classify seizures or epilepsy syndromes but one should not base the diagnosis of a single seizure or epilepsy mainly on EEG findings if good historical data are lacking.


How confident are we of the diagnosis epilepsy? Clearly, the diagnosis of a single seizure or epilepsy is subjective and will be subject to interobserver disagreement and inaccuracy. The misdiagnosis rate of 5% for the diagnosis of epilepsy found in our childhood study, must be considered an absolute minimum (2). The diagnosis was made by a panel of three experienced pediatric neurologists who discussed all patients, events could be classified as uncertain and the accuracy of the diagnosis was evaluated according to the wait and see policy. If there were no recurrences, there was no reason to change the initial diagnosis. The misdiagnosis rate of at least 23% in a British population-based study may reflect general practitioners and pediatricians without special training in epilepsy playing a central role in diagnosis and treatment (1,4,21). More specialized physicians do better. Neurologists (mistake rate 5.6%) did better than nonspecialists (mistake rate 18.9%) in another hospital-based British study (3). So ideally, all patients suspected of having epilepsy should have an assessment by a neurologist, but this is not possible in many countries due to available resources.

The diagnosis will remain uncertain for some patients even if specialists are involved. Adopting the policy of making a diagnosis of epilepsy only when the data are beyond all doubts is unrealistic and may delay effective treatment in a many patients. The risk of a diagnostic error can be minimized by taking into account all available information, especially a good eyewitness account. A home video of an event may be of great value. The previous medical history may contribute. One might argue that an EEG is indicated only if the eyewitness account has led to a reasonable suspicion of seizure(s). When events are uncertain on clinical grounds, an EEG can result in misdiagnosis due to overinterpretation, the finding of nonspecific abnormalities or the presence of (nonrelevant) epileptiform discharges that can occur in patients without epilepsy.

All these elements may be subject to interobserver variation and inaccuracy. If all the information fits into a clear pattern, a diagnosis of epilepsy can be made. If not, efforts should be made to make the information as complete as possible. When there is doubt about the diagnosis of epilepsy, patients should be classified as having an uncertain diagnosis (2,4,24). If uncertain events occur frequently, referral to a tertiary center for a second opinion or long term EEG-video monitoring may prevent a misdiagnosis of epilepsy. When the diagnosis is uncertain, prescribing AEDs with the hope that the response will clarify the diagnosis may harm the patient because of the psychosocial and socioeconomic consequences of the diagnosis, and the possible side effects of AEDs. Moreover, other important diagnoses such as a cardiac arrhythmia may be overlooked. If a patient with diagnosed epilepsy continues to have seizures after treatment with two AEDs it may be useful to reevaluate the diagnosis and classification of the seizures or epilepsy syndrome and consider tertiary center referral.


The diagnosis of a first epileptic seizure or epilepsy is subjective and subject to interobserver variation and inaccuracy. This cannot be prevented completely in our everyday care, but being aware of this problem, adopting a systematic, careful approach to the diagnosis, reassessment if AEDs fail, and timely referral to a tertiary center may be helpful to prevent a misdiagnosis of epilepsy.