Epilepsy is generally defined as a condition in which an individual tends to experience recurrent unprovoked seizures (11). A person with only one unprovoked seizure does not, by definition, have epilepsy but clearly differs from the general population in terms of risk to develop the illness.

Several studies have attempted to evaluate predictors of recurrence after a first seizure. Although estimates of the total proportion of patients who will experience recurrent seizures vary substantially across studies, much of the variation
seems related, in part, to differences in methodology. Prospective follow-up studies of individuals identified at their first seizure suggest that only approximately 25% experience a second episode within the next 2 years (12,13); this contrasts considerably with the overall recurrence risk of 70% or more reported in retrospective studies (14). Also influencing the reported variation is the heterogeneous nature of clinical epilepsy. For example, when several factors were assessed in a single study, recurrence risk at 2 years varied from less than 15% in those with no identified risk factors to 100% in those with a combination of two or more risk factors (Table 9.1) (12). Although multiple factors appear to influence recurrence risk after a first unprovoked seizure, and not all studies have identified the same predictors, a meta-analysis of studies performed through 1990 suggests some consistency in predictors and in recurrence risk after controlling for these factors (15).

A prior neurologic insult, such as neurologic deficits from birth (mental retardation, cerebral palsy), is the most powerful and consistent predictor of recurrence after a first seizure (12,13,15, 16, 17, 18). Moreover, the likelihood of a second seizure is increased by partial seizure type, abnormal electroencephalogram (EEG) (in some studies specific epileptiform EEG patterns) (12,13,16,18,19), prior acute seizures including febrile seizures (12,13), status epilepticus or multiple seizures at the index episode (12,20), and Todd’s paralysis (12,13).

In these descriptive studies, the risk for seizure recurrence has been consistently higher in patients for whom antiepileptic drugs (AEDs) have been prescribed (but not necessarily used by the patient), even after controlling for the other identified risk factors (12,13). Results from such descriptive studies may be confounded by poor compliance or failure to achieve therapeutic doses. It is reassuring, although not surprising, that in two randomized clinical trials (21,22), use of AEDs in doses to maintain serum levels in the therapeutic range was associated with a reduction in the proportion of patients who experienced seizure recurrence after a first seizure.

Individuals with multiple unprovoked seizures (epilepsy) differ clinically from those with one unprovoked seizure, and the risk for additional seizures is substantial (11). Even so, approximately 75% of all patients with a diagnosis of epilepsy enter remission (i.e., become seizure free for 5 or more years) (2). More than one half of those who enter remission do so during the first year after diagnosis (Fig. 9.1) (2).

Although predictors of seizure remission at the time of diagnosis of epilepsy may vary from those identified in association with seizure recurrence after a first seizure, some consistency is evident. Features predicting remission at initial diagnosis include young age at onset, young age at diagnosis, generalized-onset seizures, normal neurologic examination, and idiopathic or cryptogenic cause (2,23, 24, 25, 26, 27, 28). Few multivariate analyses have been performed, but these factors probably are independent. In one multivariate analysis of data from 613 children with newly diagnosed epilepsy, age (onset between 5 and 9 years of age) and a generalized idiopathic syndrome were associated with remission (29).

Conversely, a known cause, partial seizure type, neurologic deficit from birth (mental retardation or cerebral
palsy), occurrence of a primary or secondarily generalized seizure, and abnormal EEG findings—particularly a generalized spike-and-wave pattern (18,26)—are linked to a reduced likelihood of remission. The number and duration of seizures prior to diagnosis are inversely proportional to the likelihood of remission (30,31). On multivariate analysis, remote symptomatic cause, family history of epilepsy, seizure frequency, and slowing on the initial EEG were independent predictors of a reduced likelihood of remission (29). In some studies, remission was less likely in children whose epilepsy began before age 1 year, but when syndrome was taken into account, early age of onset was not a factor (32,33). Numerous seizures before initiation of treatment negatively impacted chances for remission in other studies (29,30).

During the course of epilepsy, the presence of multiple seizure types (e.g., generalized onset and partial onset) and frequent generalized tonic-clonic seizures portend a lower likelihood of remission (22,34). The duration of active epilepsy before achieving control has been thought to be the most important clinical predictor of remission (2,33). In children, failure to attain control in the first 3 months predicted a poor outcome (34,35), as did the use of multiple medications (36, 37, 38). In a study involving primarily adults (31), the total number of seizures in the first 6 months after treatment began was inversely correlated with likelihood of remission. If seizures are not controlled in the first year after diagnosis or if multiple medications are needed in the first year, only approximately 60% of patients can be expected to achieve remission (2,34,36). If seizures remain uncontrolled for more than 4 years after diagnosis, only approximately 10% of the initial cohort can be expected to enter remission after that time (2). Fewer than 5% of patients with epilepsy who continue to have seizures 10 or more years after diagnosis ever achieve total seizure control.

Early use of anticonvulsant medication does not influence the prognosis for seizure control. Randomized clinical trials from Kenya and Ecuador (39,40) indicate that 50% of patients with long-standing epilepsy achieve control if expeditiously treated with carbamazepine or phenobarbital. Presumably missing from this population are the individuals who enter remission shortly after onset. These cases probably represent the 20% of patients who fail to achieve remission early after onset of disease. In these studies, neither the duration of epilepsy nor the lifetime number of seizures affected the outcome.

Among the 60% to 70% of patients with epilepsy who enter long-term remission, 40% to 90% can have AEDs withdrawn without seizure recurrence (2,13,24,31,34,35, 41, 42, 43, 44, 45, 46, 47). Factors associated with the success or failure of AED withdrawal are familiar, being largely the same as those previously identified from initial diagnosis. Predictors of relapse after medication withdrawal include an abnormal examination in children (generally an indicator of mental retardation or cerebral palsy) (2,9,10,17,24,25,32, 33, 34,46,48); a presumed cause for the epilepsy (8,34,44,46, 47, 48); abnormal EEG findings at diagnosis (10,41,46,49), particularly a spike-wave pattern (43); worsening or persistence of an abnormal pattern (6,43); need for multiple medications to achieve control (44,50); and occurrence of many generalized seizures before control (24,25,39). In contrast, continued seizure freedom after medication withdrawal is predicted by control of seizures with monotherapy and with low serum drug concentrations, few seizures before control, and a brief interval between onset of seizures and initial control (2,24,33, 34,51). Seizure type is not a consistent predictor of relapse, although partial or secondarily generalized seizures are associated with a higher relapse rate (33,47). Young age at onset predicts successful AED withdrawal in studies of adults (52) and children (24,25,32,46,47). Family history, rate of AED withdrawal, and age at AED withdrawal (9,33,34) have been evaluated but have not been shown to influence seizure recurrence. Duration of seizure freedom may not predict successful withdrawal (33,47).