Electroclinical syndrome can be defined in approximately one third of children with new-onset epilepsy, and is one of the most robust predictors of outcome (Table 1). Most syndromes have a clearly defined natural history, and identification provides important information on likelihood of seizure control and chance of long-term remission. Furthermore, some syndromes have unique response to specific AEDs, with some drugs resulting in marked seizure reduction and others in significant exacerbation. Identification of a structural etiology on neuroimaging is also predictive of higher likelihood of pharmacoresistance; however, the specific pathology appears important. Studies have shown high rates of pharmacoresistance with cortical dysplasia, mesial temporal sclerosis, and dual pathology, but lower rates with encephalomalacia (Semah et al., 1998; Dhamija et al., 2011). Other reported adverse predictors of outcome are shown in Table 2 and include underlying neurologic deficit (Holowach et al., 1972; Annegers et al., 1979; Shorvon & Reynolds, 1982; Elwes et al., 1984; Brorson & Wranne, 1987; Hauser et al., 1996; Cockerell et al., 1997), partial onset seizures or multiple seizure types (Elwes et al., 1984; Brorson & Wranne, 1987; Collaborative Group, 1992; Cockerell et al., 1997; Arts et al., 1999), high initial seizure frequency (Elwes et al., 1984; Collaborative Group, 1992; Cockerell et al., 1997; Arts et al., 1999; Berg et al., 2001b; Brodie & Kwan, 2002; Camfield & Camfield, 2003; Mohanraj & Brodie, 2006) neonatal seizures (Berg et al., 2001b; Camfield & Camfield, 2003), early age at onset, or onset after 12 years (Casetta et al., 1999; Berg et al., 2001b; Camfield & Camfield, 2003), and failure to respond to the first AED (Sillanpaa, 1993; Camfield & Camfield, 1996; Kwan & Brodie, 2000). Predictors that are inconsistently associated with poorer outcome are epileptiform discharge or focal slowing on electroencephalography (EEG) (Shafer et al., 1988; Berg et al., 2001b; Camfield & Camfield, 2003; Spooner et al., 2006), status epilepticus (Sillanpaa, 1993; Casetta et al., 1999; Berg et al., 2001a; Callaghan et al., 2007), or history of febrile seizures (Dlugos, 2001; Hitiris et al., 2007).
Several models have been proposed to predict outcomes, based on combinations of the above factors. Based on a combined model utilizing data from both the Nova Scotia and Dutch studies on epilepsy, Geelhoed et al. (2005) found that accurate prediction of remission could be made in 70% of cases, with the two most significant factors being epilepsy type and age at first seizures. In another model identifying pharmacoresistance in children presenting before 3 years of age, probability of pharmacoresistance ranged from 0.01 to 0.97 based on the presence of absence of four factors—age at onset <12 months, developmental delay at diagnosis, neuroimaging abnormality, and focal slowing on initial EEG (Wirrell et al., 2012).