An open prospective study of brain energy metabolism was carried out in 20 consecutive cases of coma occurring after cardiopulmonary arrest (CPA) of various causes. Mean age was 54 years and mean duration of CPA 2.75 min, with a mean of 11 min for resuscitation. Brain energy metabolism was compared with clinical course, electroencephalogram (EEG), and neurological outcome. All the patients were given the usual intensive care : intermittent positive pressure ventilation, haemodynamic control, sedation with benzodiazepines, neuromuscular paralysis, anticonvulsants, mannitol. A clinical examination and an EEG were carried out daily, 4 h after all drugs which might interfere with these investigations or brain energy metabolism had been stopped (except for anticonvulsants). Successive measurements of the cerebral arteriovenous oxygen (C (a − v̄) o 2) and glucose (C (a − v̄) gluc) contents were made. The oxygen glucose index (OGI) was calculated according to Cohen's formula. The first set of measurements were carried out within the first 30 h after CPA. The last measurements were made a few hours before death or recovery, or stabilization of the coma. Patients were assigned to 2 groups according to their neurological outcome : group R (n = 8), patients who recovered consciousness ; group D (n = 12), patients who developed brain death or a vegetative state — Safar's cerebral performance category 4. During the first 30 h, there was no relationship between clinical course, EEG and cerebral outcome. Half of the patients died or recovered within 72 h. There was a statistically significant difference for C (a − v̄) o 2 (2.67 ± 0.40 mmol · 1 −1 vs. 1.36 ± 0.40 mmol · 1 −1 for groups R and D respectively ; p < 0.05) and OGI (127.1 ± 28.4 % for group R vs. 29.9 ± 7.3 % for group D ; p < 0.05). There was no statistically significant difference for C (a − v̄) gluc. The last cerebral metabolic measurements showed similar, but even more significant results (p < 0.001 for C (a − v̄) o 2 and OGI), with a significant difference for C (a − v̄) gluc (0.29 ± 0.03 mmol · 1 −1 vs. 1.52 ± 0.45 mmol · 1 −1 for groups R and D respectively ; p < 0.05). Patients from group D had a low C (a − v̄) o 2, a high C (a − v̄) gluc and a particularly low OGI : there was anaerobic glucose metabolism and an increase in glucose use. On the other hand, patients who recovered had high values of C (a − v̄) o 2, mostly low levels of C (a − v̄) gluc and unusually high OGI, i.e. there was an aerobic glucose metabolism, with consumption of a significant amount of non glucose substrates. All patients who had an initial and/or a final OGI > 85 % and a C (a − v̄) o 2 > 1.8 mmol · 1 −1 recovered. Patients with a poor cerebral prognosis had a final OGI < 85 % and a final C (a − v̄) o 2 < 1.8 mmol · 1 −1. However, patients with an OGI < 85 % and a C (a − v̄) o 2 > 1.8 mmol · 1 −1 had an unpredictable outcome. The study of cerebral oxygen and glucose metabolism can therefore help to establish an early cerebral prognosis. Cerebral recovery is linked to increased oxidation associated with non glycolytic metabolism, whereas deep brain damage is linked to persistent increased anaerobic glucose metabolism.