The tricyclic antidepressant, doxepin, is formulated as an irrational mixture of E (trans) and Z (cis) stereoisomers (85%: 15%). We examined the stereoselective metabolism of doxepin in vitro, with the use of human liver microsomes, recombinant CYP2D6 and gas chromatography-mass spectrometry. In human liver microsomes over the concentration range 5-1500 microM, the rate of Z-doxepin N-demethylation exceeded that of E-doxepin above 100 microM in two of three livers. Eadie-Hofstee plots were curvilinear indicating the involvement of several enzymes in N-demethylation. Coincubation of doxepin with 7,8-naphthoflavone and ketoconazole reduced the rates of N-demethylation of E- and Z-doxepin by 30-50% and 40-60%, respectively, suggesting the involvement of CYP1A and CYP3A4, whilst quinidine had little effect on N-demethylation. In contrast, doxepin hydroxylation was exclusively stereo-specific; E-doxepin and E-N-desmethyldoxepin were hydroxylated with high affinity in liver microsomes and by recombinant CYP2D6 (Km in the range of 5-8 microM), but there was no evidence of Z-doxepin hydroxylation. In 'metabolic consumption' experiments with liver microsomes (having measurable CYP2D6 activity) and initial substrate concentration of 1 microM, the consumption of E-doxepin was greater (P < 0.05, n = 5) than that of Z-doxepin. Quinidine inhibited the consumption of E-doxepin but did not affect the consumption of Z-doxepin. With N-desmethyldoxepin, quinidine inhibited the consumption of E-N-desmethyl-doxepin whereas Z-N-desmethyldoxepin appeared to be a terminal oxidative metabolite. In summary, CYP2D6 is a major oxidative enzyme in doxepin metabolism; predominantly catalysing hydroxylation with an exclusive preference for the E-isomers. The relatively more rapid metabolism of E-isomeric forms, and the limited metabolic pathways for the Z-isomers may explain the apparent enrichment of Z-N-desmethyldoxepin that is observed in vivo.