Objective: To quantify and analyze the value of expected information from a
n EEG after first unprovoked seizure in childhood. Background: An EEG is of
ten recommended as part of the standard diagnostic evaluation after first s
eizure. Methods: A MEDLINE search from 1980 to 1998 was performed. From eli
gible studies, data on EEG results and seizure recurrence risk in children
were abstracted, and sensitivity, specificity, and positive and negative pr
edictive values of EEG in predicting recurrence were calculated. Linear inf
ormation theory was used to quantify and compare the expected information f
rom the EEG in all studies. Standard test-treat decision analysis with a tr
eatment threshold at 80% recurrence risk was used to determine the range of
pretest recurrence probabilities over which testing affects treatment deci
sions. Results: Four studies involving 831 children were eligible for analy
sis. At best, the EEG had a sensitivity of 61%, a specificity of 71%, and a
n expected information of 0.16 out of a possible 0.50. The pretest probabil
ity of recurrence was less than the lower limit of the range for rational t
esting in all studies. Conclusions: In this analysis, the quantity of expec
ted information from the EEG was too low to affect treatment recommendation
s in most patients. EEG should be ordered selectively, not routinely, after
first unprovoked seizure in childhood.