We describe melphalan pharmacokinetics in 26 patients treated by isolated limb perfusion (ILP). Group A (n = 11) were treated with a bolus of melphalan (1.5 mg kg-1), and in a phase I study the dose was increased to 1.75 mg kg-1. The higher dose was given as a bolus to Group B (n = 9), and by divided dose to Group C (n = 6). Using high performance liquid chromatography (HPLC) the concentrations of melphalan in the arterial and venous perfusate (during ILP) and in the systemic circulation (during and after ILP) were measured. Areas under the concentration time curves for perfusate (AUCa, AUCv) and systemic (AUCs) data were calculated. In all three groups the peak concentrations of melphalan were much higher in the perfusate than in the systemic circulation. The pharmacokinetic advantages of ILP can be quantified by the ratio of AUCa/AUCs, median value 37.8 (2.1-131). AUCa and AUCv were both significantly greater in Group B than in Group A (P values less than 0.01, Mann-Whitney). In Groups B and C acceptable 'toxic' reactions occurred but were not simply related to melphalan levels. Our phase I study has allowed us to increase the dose of melphalan to 1.75 mg kg-1, but we found no pharmacokinetic advantage from divided dose administration.