The aim of this study was to re-examine the human hepatic metabolism of diclofenac, with special focus on the generation of minor hydroxylated metabolites implicated in the idiosyncratic hepatotoxicity of the drug. Different experimental approaches were used: human hepatocytes, human microsomes, and engineered cells expressing single human CYP (cytochromes P450). Human hepatocytes formed 3'-hydroxy-, 4'-hydroxy-, 5-hydroxy- 4',5-dihydroxy-, and N,5-dihydroxydiclofenac, as well as several lactams. Formation of 4'- and 5-hydroxydiclofenac by human liver microsomes followed a Michaelis-Menten kinetics (Km 9 +/- 1 microM; Vmax 432 +/- 15 pmol/min/mg and Km 43 +/- 5 microM; and Vmax 15.4 +/- 0.6 pmol/min/mg, respectively). Secondary metabolites were detected after incubation of 5-hydroxydiclofenac with human liver microsomes, yielding 4',5-dihydroxydiclofenac (Km 15 +/- 1 microM; Vmax 96 +/- 3 pmol/min/mg) and small amounts of N,5-dihydroxydiclofenac (non-Michaelis-Menten kinetics). Based on microsome studies and the incubations with human hepatocytes and engineered cells, we estimated that in vivo CYP2C9 would be exclusively responsible for the 4' hydroxylation of diclofenac (>99.5%) as well as 5-hydroxydiclofenac (>97%). CYP2C9 was exclusively responsible for the formation of 3'-hydroxydiclofenac. Multiple regression analysis evidenced that the rate of production of 5-hydroxydiclofenac in human microsomes followed the algorithm: 0.040 x S-mephenytoin 4'-hydroxylation + 0.083 x tolbutamide methylhydroxylation, (multiple correlation coefficient = 0.969). However, the incubation of diclofenac with cell lines expressing different human CYP suggested that 7 isoforms could be involved. Comparison of data obtained with CYP-expressing cells and human hepatocytes suggests that CYP2C8 > CYP2C19 approximately CYP2C18 > CYP2B6 are the isoforms implicated in the 5-hydroxylation of diclofenac in vivo.