OBJECTIVE Pentobarbital is standard treatment for refractory status epilepticus (SE) and is almost uniformly effective, but the morbidity of treatment and the mortality of refractory SE are high. Recurrence of SE after pentobarbital discontinuation may predict a worsened outcome. We sought to determine the optimal use of barbiturate anesthetic treatment of refractory SE. METHODS We reviewed 44 episodes of barbiturate anesthetic treatment for refractory SE in 40 patients, seeking factors predicting freedom from relapse to clinical or electrographic SE after treatment and predicting survival. RESULTS Eight of 9 patients with relapse of seizures after barbiturate treatment died, whereas only 9 of 26 with persistently controlled seizures died (p < 0.005). Both likelihood of relapse and survival correlated strongly with etiology, with 19 of 20 patients with chronic epilepsy, infections, or focal lesions having good control as compared with 2 of 9 with multiple medical problems (p < 0.001). Treatment delay did not predict a worsened outcome. Hypotension caused dose reduction but never required treatment discontinuation. Patients with more prolonged treatment and those receiving phenobarbital (PB) at the time of pentobarbital taper were less likely to relapse. CONCLUSIONS Relapse of SE after barbiturate anesthetic treatment is a poor prognostic sign and should be prevented, if possible. Etiology was the primary predictor of outcome, but more prolonged treatment and the use of PB during pentobarbital withdrawal appeared to provide protection against relapse.