Formation of 6 alpha-hydroxypaclitaxel has been described as the primary biotransformation pathway for paclitaxel in vitro and in vivo, with additional formation of two other "minor" metabolites. Using a large group (n = 49) of human liver microsomes, and P450s heterologously expressed in cell lines, our aims were to elucidate the P450s responsible for and investigate variability in paclitaxel metabolite formation. Four metabolites of paclitaxel (6 alpha-hydroxypaclitaxel, metabolites B, C and A) were formed in vitro, via CYP2C8, 3A4, 3A4 and both 2C8 and 3A4, respectively. Although 6 alpha-hydroxypaclitaxel was predominant in the majority of livers, metabolites B and C (formed by CYP3A4) were predominant in 11/49 and 2/49 livers, respectively. Predominance of metabolite B over 6 alpha-hydroxypaclitaxel was more likely in liver microsomes from donors known to be exposed to phenobarbital (P = .009), and tended to be more likely in diseased vs. normal livers (P = .047). Formation rates for 6 alpha-hydroxypaclitaxel, A, B, and C were lower in diseased liver vs. normal liver (P < .001). Rates of formation of metabolites B and C were highly correlated with each other (r2 = .91; P < .001) and with midazolam 4-hydroxylation (r2 = .87 & 0.86, respectively; P < .001). Inhibitor experiments suggest that typical CYP3A substrates/inhibitors (e.g., cyclosporin, epipodophyllotoxins) may significantly interact with paclitaxel in vivo. In a single patient in whom plasma samples were measured on two occasions, metabolite A (the dihydroxylate) was predominant, and systemic clearance of paclitaxel was lower in a course administered 1 day vs. 6 wk after a course of fluconazole therapy. We report that 6 alpha-hydroxypaclitaxel, formed via CYP2C8, is not the predominant paclitaxel metabolite in all individuals, and that CYP3A4 catalytic activity is important to overall paclitaxel metabolism in humans.