To characterize the population pharmacokinetics of amikacin in intensive care unit (ICU) patients and to analyse whether these patients show different kinetic behaviour on the basis of their clinical diagnoses. The patient population comprised 104 medical ICU patients on amikacin treatment for several presumed or documented Gram-negative infections. Four study groups were defined according to patients' clinical diagnosis: sepsis group (n = 39), trauma group (n = 20), pneumonia group (n = 21) and 'other diagnosis' group (n = 24). The pharmacokinetic parameters for amikacin in these patients were then compared. The ICU patients were found to have increased values for the amikacin volume of distribution (0.52 +/- 0.21 litres/kg), whereas total amikacin clearance expressed as a linear function of creatinine clearance was Cl (ml/min/kg) = 0.13 +/- 0.86 ClCR, which is not significantly different from other estimations reported in the literature. However, this relationship revealed statistically significant differences among the four groups of ICU patients. Moreover, the septic and trauma patients showed higher (but not statistically significant) values for the amikacin volume of distribution. The amikacin pharmacokinetic parameters obtained should allow Bayesian individualization of amikacin doses in patients admitted to medical ICUs, on the basis of their clinical diagnoses.