Tirapazamine (3-amino-1,2,4-benzotriazine-1,4-di-N-oxide; SR 259075) is a selective hypoxic cell cytotoxic agent: that is bioreductively activated in tumours to a reactive-drug free radical. Preclinically the agent has been shown to possess additive and synergistic anti-tumour activity in combination with radiotherapy and chemotherapy regimens. In the present study the pharmacokinetics and metabolism of tirapazamine were investigated in mice and patients as part of pre-clinical and phase I investigations. The objectives of this work were twofold; firstly, to evaluate retrospectively the utility of a pharmacokinetically guided dose-escalation (PGDE) strategy for tirapazamine, and secondly, to investigate if pharmacologically relevant plasma concentrations could be achieved at tolerable doses. Pharmacokinetic studies for PGDE were conducted in mice at four dose levels ranging from one-tenth of the LD10 to the LD50. The AUC at the LD10 (2932 μg ml-1 min) was used to determine a target AUC value of 1173 μg ml-1 min (equivalent to 40% of the mouse LD10 AUC) for clinical studies. A phase I study to investigate the tolerance of a single i.v. infusion of tirapazamine (once every 3 weeks) was initiated with close pharmacokinetic monitoring. The starting dose (36 mg/m2) was based on toxicity data obtained in the mouse, rat and dog. Doses were escalated by increases in the volume and duration of infusion. A retrospective analysis of the pharmacokinetic and toxicity data was then made to determine the utility of a PGDE approach. The drug exhibited a steep dose-lethality relationship in mice (LD10 294 mg/m2, LD50 303 mg/m2). The major gross toxicities were body-weight loss (15-20%), pilo-erection and hypoactivity at all dose levels. Sporadic ptosis and conjunctivitis were observed at doses of > 300 mg/m2. The plasma elimination of tirapazamine fitted a monoexponential open model, with rapid elimination from the plasma (t( 1/4 ) = 36±0.65 min) occuring at the LD10 dose of 294 mg/m2. A 10.3-fold increase in dose resulted in a 25.0-fold increase in AUG. Clinically, doses were escalated over the range of 36-450 mg/m2. Ototoxicity (tinnitus and reversible hearing loss) was dose-limiting at 450 mg/m2 and the MTD was 390 mg/m2 for this schedule. Pharmacokinetic analyses in patients revealed that the elimination of tirapazamine in patients was generally bi-phasic, with low inter-patient variability being found in clearance. A 12.5-fold increase in dose resulted in a 19.0-fold increase in AUC. There was good quantitative agreement in metabolite formation between mice and humans with respect to the two- and four-electron bioreductive metabolites. AUC values recorded for tirapazamine at the MTD of 390 mg/m2 (range 1035-1611 μg ml-1 min) were similar to the target AUC in mice. Importantly, these levels are consistent with the levels required for radiation-dose enhancement and effective combination with cisplatin in mice. Given (a) the similarities in plasma pharmacokinetics and metabolism observed at the target AUC/MTD in mice, rats, dogs and humans, (b) the similar degree of plasma protein binding seen between species and (c)the relatively low inter-patient variability noted in drug clearance, a successful PGDE approach should have been feasible. The results also indicate that potentially therapeutic levels of tirapazamine are achievable in patients at tolerable doses.