Aims Using human liver microsomes and heterologously expressed human enzymes, we have investigated the involvement of CYPs 1A2, 2C9, 2C19, 2D6 and 3A4 in the N-demethylation of amitriptyline (AMI), with a view to defining likely influences on its clinical pharmacokinetics. Methods The kinetics of formation of nortriptyline (NT) from AMI were measured over the substrate concentration range 1-500 mm, using liver microsomes from four extensive metabolisers (EM) and one poor metaboliser (PM) with respect to CYP2D6 activity. Results The data were best described by a two-site model comprising a MichaelisMenten function for a high affinity site and a Hill function for a low affinity site. The activity at the low affinity site was eliminated by triacetyloleandomycin and ketoconazole, selective inhibitors of CYP3A4, such that the kinetics were then described by a two-site model comprising two Michaelis-Menten functions. A further decrease in activity was associated with the addition of the CYP2C9 inhibitor sulphaphenazole such that the residual kinetics were best described by a single Michaelis-Menten function. The addition of quinidine, a selective inhibitor of CYP2D6, along with triacetyloleandomycin and sulphaphenazole produced an additional decrease in the rate of NT formation in all but the PM liver, but did not completely eliminate the reaction. The remaining activity was best described by a single Michaelis-Menten function. Inhibitors of CYP1A2 (furafylline) and CYP2C19 (mephenytoin) did not impair NT formation. Microsomes from yeast cells expressing CYP2D6 and from human lymphoblastoid cells expressing CYP3A4 or CYP2C9-Arg N-demethylated AMI, but those from cells expressing CYPs 1A2 and 2C19 did not. Conclusions We conclude that CYPs 3A4, 2C9 and 2D6 together with an unidentified enzyme, but not CYPs 1A2 and 2C19, mediate the N-demethylation of AMI. Thus, the clinical pharmacokinetics of AMI would be expected to depend upon the net activities of all of these enzymes. However, the quantitative importance of each isoform is difficult to predict without knowledge of the exposure of the enzymes in vivo to AMI.