Objectives: The vagus nerve has important immunological and anti-inflammatory actions that might be relevant to the beneficial effects of vagus nerve stimulation (VNS). Therefore, we conducted an exploratory study on VNS effects on cytokine levels in plasma and cerebrospinal fluid of children suffering from refractory epilepsy. Moreover, as predictors of the response are lacking, we also aimed to determine if cytokine changes predict the clinical response. Methods: VNS was performed according to a randomized double-blind design: plasma levels were compared between patients who received 20 weeks of high output (therapeutic) (n = 21) or low output (active control) stimulation (n = 20). Thereupon, all patients received high output stimulation for another 19 weeks; levels during this period were compared to baseline. Interictal interleukin-1β, interleukin-6, and interleukin-10 were determined by ELISA. Results: No significant changes were found between high and low output groups and between the last 19 weeks of stimulation and baseline. Changes in interleukin-1β correlated with improved IQ (τ = 0.42, p < 0.01). Lower baseline plasma levels of interleukin-6 were associated with more seizure frequency reduction [R2 = 0.105 (1, 35), p = 0.050]. Conclusion: Interictal cytokine levels were not altered by VNS but baseline interleukin-6 predicted the clinical response. In the future, patient selection may be aided by determination of the cytokine profile of the patient.