In 1987 and 1988, we carried out a prospective study of patients older than 10 years with nontraumatic coma in the intensive care units of Columbia-Presbyterian Medical Center, New York, NY. Of 188 patients with Glasgow Coma Scale (GCS) determinations within 72 hours, 61% were dead or in persistent coma by 2 weeks from onset. Age, sex, and ethnicity did not influence outcome. The 2-week outcome for patients with initial GCS of 3 to 5 was 14.8% awake; 85.2% were dead or in persistent coma. For the GCS 6 to 8 group, 53.1% were awake and 46.9% were dead or in persistent coma. Hypoxic or ischemic coma had the worst 2-week outcome (79% dead or comatose); coma caused by metabolic disease or sepsis (68%), focal cerebral lesions (66%), and general cerebral diseases (55%) were intermediate, while drug-induced coma had a favorable outcome (27% dead or comatose). The independent predictors of 2-week outcome were the first GCS and drug-induced coma. The predicted probability of waking at 2 weeks was eight times better for drug-induced coma than other causes when GCS was held constant. Patients with an initial GCS score of 6 to 8 were seven times more likely to waken than those with a score of 3 to 5. The motor subscore alone was a significant independent predictor of 2-week outcome. Modification of coma score to include etiology may give more accurate predictions of 2-week outcome after nontraumatic coma.