Objective: To evaluate the prognostic value of middle-latency somatosensory evoked potential (MLSEP) in comatose patients with acute severe supratentorial stroke, considering both unfavourable outcome and death. Methods: Eighty-eight patients with acute severe supratentorial stroke underwent MLSEP, short-latency somatosensory evoked potential (SLSEP), Glasgow Coma Scale (GCS) and cerebral computed tomography (CCT) within 1 week from onset. MLSEP and SLSEP were recorded in 25 normal controls. All patients were evaluated with two criteria of outcome as unfavourable outcome (modified Rankin Scale 4-6) and death 6 months after onset. N60 of MLSEP predictive value was compared with N20 of SLSEP, GCS and CCT. Results: Sixty-seven patients (76.1%) suffered from cerebral infarction; and 21 patients (23.9%) suffered from intracerebral haemorrhage. Seventy-one patients (80.7%) had unfavourable outcomes and 39 patients (44.3%) died. The peak latencies of MLSEP were prolonged and some waves of MLSEP were absent in stroke patients, and the proportion of absent waves in lesion-ipsilateral MLSEP was higher than in contralateral MLSEP. By using the prognostic authenticity analysis of predictors, the lesion-ipsilateral absence of N60 showed the highest sensitivity for unfavourable outcome (97.2%, confidence interval (CI): 89.3-99.5%) and death (100%, CI: 88.8-100%), which was superior to GCS, CCT and N20. Bilateral absence of N60 showed a high specificity of 100% for unfavourable outcome, which was as good as bilateral absence of N20. However, it showed a specificity of 89.8% (CI: 77.0-96.2%) for death, not as good as bilateral absence of N20 (98%, CI: 87.8-99.9%). The false positive rate of lesion-ipsilateral absence of N60 for unfavourable outcome and death was 12.7% (CI: 6.6-22.5%) and 50.6% (CI: 39.2-62.0%), respectively, and that of bilateral absence of N60 was 0 (CI: 0-12.3%) and 14.3% (CI: 5.4-31.0%), respectively. Conclusions: We confirm the high predictive value of MLSEP in severe stroke. MLSEP showed higher sensitivity than SLSEP for predicting unfavourable outcome and death. Combined MLSEP with SLSEP results produced even greater predictive value. Significance: The combination of MLSEP and SLSEP would increase the sensitivity and maintain the high specificity not only for predicting outcome in coma after cardiopulmonary resuscitation but also after severe stroke. (C) 2011 Published by Elsevier Ireland Ltd. on behalf of International Federation of Clinical Neurophysiology.