Sensory evoked potentials are markedly changed in patients with fulminant liver failure. It is unknown, however, whether serial recordings of sensory evoked potentials provide useful prognostic informations for patient management in fulminant liver failure. Ninety recordings of bilateral median nerve-stimulated short- and long-latency sensory evoked potentials were performed in 25 patients with fulminant liver failure (9 patients spontaneously recovered and survived, 8 patients were referred to emergency liver transplantation and 8 patients died). In all nine survivors the cortical long-latency sensory evoked potential N70 peak was constantly detectable between 74 and 162 ms. In all eight patients who subsequently underwent liver transplantation and in seven of eight patients who died, loss of the N70 peak developed during the course of fulminant liver failure. In 4 of 15 patients who were selected for liver transplantation according to the King's College criteria, the N70 peak was constantly detectable. All four transplantation candidates spontaneously recovered and survived without transplantation. In contrast, eight patients never did fulfill the criteria for liver transplantation. Five of them with constantly detectable N70 peak recovered spontaneously. However, in the remaining three patients loss of the N70 peak developed, and they ultimately died. The probability of correct outcome prediction by sensory evoked potentials is superior to that based on clinical criteria alone (0.96 vs. 0.72). Therefore serial recording of sensory evoked potentials may help identify (a) a subgroup among liver transplantation candidates who may spontaneously recover without transplantation and (b) a subgroup of patients with severe brain dysfunction who should undergo early liver transplantation even though they do not fulfill King's College criteria.