Objective: The aim of this study was to investigate whether serial serum neuron-specific enolase (NSE) can be used to predict neurological prognosis in patients remaining comatose after cardiopulmonary resuscitation (CPR). Design: Observational cohort study. Clinicians were blinded to NSE results. Setting: Eighteen-bed general ICU. Patients: Comatose patients admitted to the ICU after CPR. Interventions: Serum NSE was measured at admission and daily for 5 days. Measurements and results: Patients received full intensive treatment until recovery or until absence of cortical response to somatosensory evoked potentials more than 48 h after CPR proved irreversible coma. Of the 110 patients included (mean GCS at ICU admission 3, range 3-9), 34 regained consciousness, five of whom died in hospital. Seventy-six patients did not regain consciousness, 72 of whom died in hospital. Serum NSE at 24 h and at 48 h after CPR was significantly higher in patients who did not regain consciousness than in patients who regained consciousness (at 24 h: median NSE 29.9 μg/l, range 1.8-250 vs 9.9 μg/l, range 4.5-21.5, P <0.001; at 48 h: median 37.8 μg/l, range 4.4-411 vs 9.5 μg/l, range 6.2-22.4, P = 0.001). No patient with a serum NSE level >25.0 μg/l at any time regained consciousness. Addition of NSE to GCS and somatosensory evoked potentials increased predictability of poor neurological outcome from 64% to 76%. Conclusions: High serum NSE levels in comatose patients at 24 h and 48 h after CPR predict a poor neurological outcome. Addition of NSE to GCS and somatosensory evoked potentials increases predictability of neurological outcome.