In this post hoc analysis, individual seizure counts from four double-blind trials of adjunctive treatment with levetiracetam were analyzed by non-linear mixed-effects modeling (NONMEM). First, a model was fitted to the individual count data assuming a Poisson distribution, in order to classify the patients as either improving or deteriorating from baseline. In the second stage, the dose-response relationship in improving patients was determined by fitting the data to an E(max) model including a placebo effect. The percentage of improvers was 59% on placebo and 73%, 74%, 77% and 73% on levetiracetam 1, 2, 3 and 4g/day, respectively. The ED(50) of 1408mg/day was close to the current WHO Defined Daily Dose of levetiracetam (1500mg). The maximum recommended dose of 3000mg/day was predicted to reduce seizures by >or=90% in 10% of improving patients. Age, gender, body weight, race, and number of concomitant antiepileptic drugs neither affected the percentage of responders nor the extent of change in seizure frequency from baseline.