Search strategy, selection criteria, and data abstraction

The current study was initiated in response to a regulatory request from the Pediatric Committee of the European Medicines Agency as part of regulatory guidance in the pediatric approval process for the antiseizure drug perampanel. Electronic searches of EMBASE, Medline, and the Cochrane Central Register of Controlled Trials were conducted for controlled trials in the treatment of PGTCS published from 1970 to 2015 using predefined criteria (text e-1, doi.org/10.5061/dryad.cg24gt6). In this meta-analysis, we interpret the term primary generalized tonic-clonic seizures to refer to generalized tonic-clonic seizures that occur in the context of the idiopathic (presumed genetic) generalized epilepsies. In doing so, we also accept that purely generalized tonic-clonic seizures may occur outside of the widely recognized epilepsy syndromes and also that seizures that appear to be generalized at onset may also occur in the context of focal epilepsies, where rapid secondary generalization may occur. The use of this term was meant to restrict the analysis to a more homogeneous population, although in clinical practice we concede this may be difficult to apply in all cases.

The initial online database search for trials on epilepsy and tonic-clonic seizures identified 19,485 citations (figure 1). To minimize the risk of bias in individual studies, only publications that presented results from controlled, randomized, double-blind trials were included. Duplicate titles resulting from searches of more than 1 database and slight variations in the entries of titles or authors in the databases were removed. Randomized clinical trials were selected (302) and conference proceedings were excluded. A neurologist (A.L.) and a statistician (J.W.) reviewed 262 unique abstracts by title, key words, and content. Full-text articles were obtained for 39 abstracts for controlled, randomized clinical trials potentially relating to PGTCS. Articles were evaluated by A.L. and D.R.N. and included in the meta-analysis if the data were based on randomized, placebo-controlled studies on PGTCS, and contained efficacy evaluations expressed as change in percent seizure frequency or ≥50% responder rates. A neurologist (A.L.) then reviewed the selected articles and confirmed previous exclusions.

A total of 7 published trials for adjunctive therapy for PGTCS were eligible for the meta-analysis. The quality of each trial was assessed using a published instrument to evaluate the methodology and clinical relevance.⁶ Quality scores ranging from 0 to 1 were assigned for each trial.⁶ The average quality score for the 7 trials used in this analysis was 0.74 with an SD of 0.1 and range of 0.51 to 0.81. To minimize the risk of bias across studies, all relevant data within the 7 publications were used and there was no selective reporting within studies, an approach that avoided selection bias.