Hematologic toxicity of carboplatin is largely dependent on its pharmacokinetics. Its seems likely that therapeutic efficacy is also related to plasma drug exposure. Dosage adjustments based on isotopic determination of glomerular filtration rate have been proposed but their ambulatory use is not conceivable. A population pharmacokinetic study was undertaken to determine a relationship between patient characteristics and carboplatin clearance. Plasma carboplatin pharmacokinetics were determined as ultrafilterable platinum in 70 patients (23 to 84 years old) treated with different combination regimens including carboplatin at doses ranging from 184 to 950 mg (1-hr i.v. infusion) for various tumor types. Data were analysed using Nonlinear Mixed Effects Model (NONMEM). The data from 34 patients (46 cycles) were used to obtain the most predictive formula for the carboplatin clearance (ml/min): [formula: see text] The obtained formula was prospectively evaluated with the data from 36 other patients (43 cycles) and compared to other methods available to predict carboplatin clearance. Prospectively this formula predicted the clearance with a good precision (median absolute percent error of 10% range 0-30%) and minimal bias (median percent error: 2% range--25-30%). This method of prediction was as accurate as the one which requires the glomerular filtration rate to be measured by 51Cr-EDTA injection. This formula should allow to individualize very easily the carboplatin dosage in adults by multiplying the calculated carboplatin clearance by the area under the curve desired for administration.