• We evaluated the use of Bayesian forecasting for gentamicin therapy in outborn infants weighing 1000 g or less irrespective of postnatal age. Dosages were individualized using a computer program, guided by early serum gentamicin assays after a loading dose and a database of population kinetics. Steady-state gentamicin levels achieved were compared with those from a regimen based on guidelines. A total of 26 gentamicin courses were individualized in 19 infants of 22 to 33 weeks' gestation, weighing 500 to 1000 g at 1 to 41 days of age. All steady-state trough levels were between 1 and 2.4 mg/L; peak levels were between 4.4 and 9.3 mg/L. The 95% confidence intervals were in almost identical ranges. The prevalence of toxic and suboptimal trough levels was less when compared with that of 23 gentamicin courses based on guidelines in 17 control infants. We conclude that early individualized gentamicin dosage over a range of postnatal age is a practical alternative and serum level distributions appear superior.