A cooperative prospective study of consecutive cases of carbamazepine overdose was conducted to determine if serum levels were predictive of toxicity and if risk factors such as age, chronic exposure, or previous disorder or cardiovascular disease could be used as prognostic indicators. Seventy-three consecutive cases were collected from two regional certified poison control centers from January 1989 to August 1989. There were 25 exposures in children less than 6 yrs., 11 exposures in adolescents, and 37 exposures in adults. Ten adult cases and one adolescent case were excluded from the study due to the presence of coingestants or inadequate information. Peak measured serum levels ranged from 0.3 to 56 mcg/ml. Using the presence of coma, seizure activity or respiratory depression requiring mechanical ventilation as measures of toxicity, we found poor correlation between rising serum levels of carbamazepine and toxicity. Increased serum levels of carbamazepine did appear to correlate with increased hospital stay, but not with ICU stay. History of a seizure disorder appears to pose increased risk of a seizure in carbamazepine overdose. In this series chronic exposure to carbamazepine did not appear to increase the risk of coma or respiratory depression for a given toxic serum level and may add some protective effect. Serum levels below 40 mcg/ml do not appear to accurately predict the severity of toxicity. Cardiac conduction defects were rare (one child). Anticholinergic findings, as evidence by decreased bowel motility and sinus tachycardia were common. Previous cardiovascular disease and age did not appear to be important prognostic indicators.