The accuracy of the diagnosis of paroxysmal events in children

Patients and methods.

The prospective multicenter hospital-based DSEC started in 1988. After informed consent, we enrolled consecutively all children aged 1 month to 16 years who were referred because of a possible single seizure or epilepsy. Most children were referred directly by their general practitioner or by the pediatricians of the participating hospitals (who routinely refer all children with possible seizures to the departments of pediatric neurology). Some were first seen in the emergency department. We excluded children with only neonatal, febrile, or other acute symptomatic seizures as well as children who were referred from other hospitals for a second opinion.

The attending pediatric neurologist completed an extensive questionnaire on the description of the events including postictal signs, possible provoking factors, medical history, and family history. In addition, the descriptions of the episodes according to the letters to the general practitioners were made available to the panel judging the diagnoses (see below). We performed a standard EEG in each child. If this did not show epileptiform discharges, a recording after partial sleep deprivation or, in small children, during a daytime nap was made. Details on the recording and classification of the EEG have been described elsewhere.8 A brain CT scan was performed in all children unless anesthesia was required or the child had idiopathic generalized epilepsy with absences. Seventy-two percent of the children with one seizure and 66% of the children with multiple seizures had a CT scan. In former publications on the DESC, more extensive information concerning the results of imaging can be found.6,7⇓

A panel of three of four pediatric neurologists with at least 10 years of experience in pediatric epilepsy (H.S., A.C.B.P., O.F.B., and W.F.M.A.) judged whether the description of the ictal events fulfilled the predefined descriptive diagnostic criteria for seizures as adapted from an earlier Dutch study in adults9 (table E1 on the Neurology Web site). To prevent bias, none of the pediatric neurologists was allowed to judge his patients, but only those who were treated by one of the other panel members. These pediatric neurologists in 1988 founded the DSEC and have been involved in it since then. They discussed all 881 children referred during monthly sessions. Unanimous decisions were reached in all cases. The 121 children in whom another diagnosis was made were excluded. The study group consisted of 760 children: mean age 5.4 years, median age 4 years (25 to 75% range: 2 to 9 years); 367 were boys. A flow chart of all children in the study and their subgroups is presented in the figure. For a single event, the diagnosis was based exclusively on the description of the event without knowledge of the results of the EEGs or other ancillary studies. We considered a single unclear event in a child with an abnormal EEG not as an epileptic seizure, as epileptiform discharges are present in some children without any clinical seizures. In case of multiple events, however, the results of the EEGs were considered if the panel agreed based on their clinical experience that the events were suspect for seizures, although the criteria were not completely satisfied.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Flow chart of the 881 children with one or more possible seizures.

In the DSEC we followed up all children with a diagnosis of epilepsy for 5 years and all children with a single seizure for 2 years to assess the prognosis.6,7⇓ At 2 years, the accuracy of the original diagnosis was assessed. In an attempt to include all children with epilepsy in the DSEC, children with unclear events (including children with presumed pseudo-seizures) were followed for 1 year to assess whether new episodes might yield firm evidence for a definite diagnosis. An unclear event was defined as a paroxysmal event not considered as an epileptic seizure, but without another obvious explanation. After each new event, the patient was reevaluated and the diagnosis reconsidered. Point estimates regarding false-positive or false-negative diagnoses are given with their 95% CIs. The CIs were calculated according to the efficient-score method (corrected for continuity).10

Results.

Discussion.

In this study rigid diagnostic standards were applied for the diagnosis of a single seizure or epilepsy in childhood. The diagnosis was based on consensus between the members of a panel of experienced pediatric neurologists, using predefined diagnostic criteria. Unfortunately, a gold standard to confirm or refute the diagnosis of a seizure or of epilepsy is lacking. Instead, we tried to disprove the original diagnosis by assessing whether new episodes during follow-up might lead to another diagnosis. Using this approach, we found a false-positive rate for the initial diagnosis “single seizure” of 0% and for “epilepsy” of almost 5%. It should be noted that this method probably underestimates the number of incorrect diagnoses, because the diagnosis will not change if the child does not have a recurrence of the events during follow-up. Moreover, the low false-positive rate may be explained by the fact that the panel tended to prefer the diagnosis unclear and not seizure or epilepsy in case of any doubt or disagreement between the panel members.

In addition, we studied the fate of the children with an unclear diagnosis at enrollment. The false-negative rate was 7.4% for children recruited after a single unclear event and 5.6% for children enrolled after more unclear events. In other words, irrespective of the fact that a child was enrolled after one or more unclear events, one of every 15 children was not recognized to have epileptic seizures. However, none of the children experienced any harm due to this delay.

Most of the children in whom the diagnosis seizures was changed during follow-up, as well as the children with unclear events in whom a diagnosis became obvious during follow-up, turned out to have a harmless paroxysmal disorder (see table 3). The majority was not harmed by unnecessary delay of medical treatment. The only exception was the child with TIAs based on moyamoya disease diagnosed initially as epilepsy.

In view of the rigid standards we used in the diagnostic process (diagnosis by a unanimous decision of a panel of three experienced pediatric neurologists with use of predefined criteria), the figures found in our study may be a minimum estimate for the errors in everyday clinical and research situations. This should be kept in mind when judging studies on the prognosis of single seizures and epilepsy in children. We recommend that in this type of research, the initial diagnosis should be reconsidered after sufficient follow-up. From a clinical point of view, however, a false-negative diagnosis of epilepsy is probably less harmful for the patient than a false positive. A conservative approach in children with paroxysmal events of uncertain nature seems, as before, warranted.