At present, the determination of steady-state trough serum/plasma concentrations of clozapine is considered a useful tool for the clinical management of schizophrenic patients treated with this drug. In a previously published study, it was indicated that only plasma should be used to avoid a significant underestimation of clozapine and norclozapine concentrations; however, a formal evaluation of this topic has still not been made, and a consensus on the use of plasma or serum for therapeutic clozapine monitoring may be desirable. Paired samples of serum and plasma (K3EDTA solution contained in Vacutainer tubes) were obtained from 40 schizophrenic patients, and clozapine and norclozapine concentrations were determined by high-performance liquid chromatography. For the parent drug and its metabolite, serum concentrations were higher than in plasma (approximately 7%), although the correction of plasma concentrations in function of hematocrit values reduced this difference to 3%. High correlation coefficients were found between the serum and uncorrected or corrected plasma clozapine concentrations (r = 0.996, P < 0.001), with clinically acceptable differences between the means and standard error of the estimate and consequently with transferability of the results. The clozapine and norclozapine concentrations in five lithium heparin-containing plasma samples (371.9 +/- 226.7 ng/mL and 217.9 +/- 113.1 ng/mL) were analogous to the corresponding hematocrit-corrected EDTA-containing plasma values (374.4 +/- 225.4 ng/mL and 223.5 +/- 115.2 ng/mL), with correlation coefficients of r > or = 0.998 (P < 0.001). Serum or plasma samples may be used for the therapeutic monitoring of clozapine, and no practical advantages have been found with regard to the stability of the drug or imprecision obtained by using either type of biological matrix.