A multivariable prediction model of a major treatment response for focal-onset seizures: A post-hoc analysis of Phase III trials of perampanel

Gregory L. Krauss, Elinor Ben-Menachem, Robert T. Wechsler, Anna Patten, Betsy Williams, Antonio Laurenza, Manoj Malhotra

Research output: Contribution to journalArticlepeer-review


Objective: Although 50 % reduction in seizure frequency is a common efficacy endpoint in clinical trials of antiepileptic drugs (AEDs), 75 % or greater reductions may be required to improve patients’ health-related quality of life. Identification of clinical factors that are associated with high responder rates may help to inform clinicians on which patients may optimally benefit from treatment. We evaluated potential predictive factors for achieving major treatment responses (≥75 % reduction in seizure frequency per 28 days from study baseline) in patients with drug-resistant focal-onset seizures, with/without focal to bilateral tonic-clonic (FBTC) seizures in perampanel trials designed for regulatory approval. Methods: Univariate analyses using logistic regression were performed using data from three double-blind, placebo-controlled Phase III studies of adjunctive perampanel (Studies 304 [NCT00699972], 305 [NCT00699582], 306 [NCT00700310]), and their open-label extension study (OLEx; Study 307 [NCT00735397]). For the double-blind studies, baseline seizure frequency, number of baseline AEDs, baseline seizure type, baseline concomitant enzyme-inducing AEDs (EIAEDs), baseline carbamazepine, lamotrigine, or valproic acid, age at diagnosis, time since diagnosis, etiology, and perampanel plasma concentration were included individually with study treatment. The same factors were included for the OLEx analysis except for plasma concentration and treatment. Variables found to be significant predictors for a major treatment response in univariate analyses were subsequently included in multivariable analyses using backwards and forwards selection. Results: In the double-blind studies, 175/1374 patients had a major response to placebo (n = 25) or perampanel (n = 150). The best predictors of a major treatment response in multivariable models with forwards and backwards selection were: the presence of FBTC seizures during baseline (P = 0.0002), higher perampanel plasma concentration (P < 0.0001), older age at diagnosis (P = 0.0024 and 0.0045, respectively), and lower baseline seizure frequency (P = 0.0364 and 0.0127, respectively). In the OLEx, 217/1090 patients had a major treatment response. The best predictors of a major treatment response in the final multivariable model, regardless of backwards or forwards selection, were a lower baseline seizure frequency (P = 0.0022), the absence of focal impaired awareness seizures during baseline (P = 0.0011), the presence of FBTC seizures during baseline (P = 0.0164), lower number(s) of baseline AEDs (P = 0.0002), the absence of EIAEDs during baseline (P = 0.0059), an older age at diagnosis (P = 0.0054), and absence of structural etiologies (P = 0.0138). Significance: These analyses of placebo-controlled and long-term extension trial data identified a number of potential predictive factors for patients with focal-onset seizures achieving a major treatment response. These factors may help guide clinicians when predicting a patient's response to treatment and optimizing individual treatment regimens.

Original languageEnglish (US)
Article number106649
JournalEpilepsy Research
StatePublished - Aug 2021


  • Antiepileptic drugs
  • Epilepsy
  • Focal-onset seizures
  • Perampanel
  • Predictive factors
  • Treatment response

ASJC Scopus subject areas

  • Neurology
  • Clinical Neurology


Dive into the research topics of 'A multivariable prediction model of a major treatment response for focal-onset seizures: A post-hoc analysis of Phase III trials of perampanel'. Together they form a unique fingerprint.

Cite this