1. Seven patients clinically diagnosed as being hypersensitive to carbamazepine and one patient hypersensitive to both carbamazepine and oxcarbazepine have been identified. They have been compared with a control group (hereafter referred to as 'control subjects') comprising five patients on chronic carbamazepine therapy without adverse effects and 12 healthy volunteers who have never been exposed to anticonvulsants. 2. An in vitro cytotoxicity assay employing mononuclear leucocytes as target cells has been used first, to determine the ability of 10 different human livers to bioactivate carbamazepine to a cytotoxic metabolite, and secondly, to compare the cell defences of carbamazepine-hypersensitive patients and control subjects to oxidative drug metabolites generated by a murine microsomal system, using a blinded protocol. 3. With human liver microsomes, the metabolism-dependent cytotoxicity of carbamazepine increased with increasing microsomal protein concentration. At a protein concentration of 2 mg per incubation, the cytotoxicity of carbamazepine with human liver microsomes (n = 10 livers) increased from 7.2 +/- 0.8% (baseline) to 16.4 +/- 2.1% (with NADPH; P = 0.002). 4. In the presence of phenobarbitone-induced mouse microsomes and NADPH, the mean increase in cytotoxicity above the baseline with carbamazepine was significantly greater (P less than 0.001) for the cells from the carbamazepine-hypersensitive patients (7.9 +/- 0.8%) than from control subjects (2.6 +/- 0.3%). 5. In the presence of phenobarbitone-induced mouse microsomes and NADPH, there was no significant difference in cytotoxicity between the cells from carbamazepine hypersensitive patients and from control subjects in the presence of either phenytoin or oxcarbazepine.(ABSTRACT TRUNCATED AT 250 WORDS)