Thirty-seven premature neonates with suspected infection at Neonatal Intensive Care Unit (NICU), Phramongkutklao Hospital have been given amikacin as an empirical treatment. Dosage of amikacin are 18 mg/kg every 48 hours, 16 mg/kg every 48 hours and 15 mg/kg every 24 hours for neonates with gestational age not greater than 30 weeks, 31 - 33 weeks and 34 - 36 weeks, respectively. Medications have been administered by intravenous infusion for 30 minutes. For each neonates, two amikacin serum levels have been obtained at steady state. Peak concentration has been measured at 30 minutes after complete infusion. Second concentration has been measured at either 18th hours or 36th hours after administration depend on dosing interval of each patient. Trough concentration has been derived from calculation. Pharmacokinetic parameters of amikacin also obtained from calculation. Mean peak concentration at steady state of neonates with gestational age not greater than 30 weeks, 31-33 weeks and 34-36 weeks are 28.49 +- 8.63, 24.22 +- 5.99 and 24.91 +- 5.73 microgram/ml, respectively. Mean trough concentration at steady state are 1.65 +- 1.28, 1.13 +- 2.11 and 2.01 +- 1.12 microgram/ml, respectively. Mean peak and trough concentration are not significantly different between groups (p>0.05). Mean peak concentration of every groups are in desired therapeutic range (20-30 microgram/ml) as mentioned in Neofax 2001. Mean trough concentration of every groups are not greater than desired range (2-5 mg/ml). Elimination rate constant and clearance of amikacin are significantly different between groups and are also good proportional correlation with gestational age (p<0.05). If extreme value has been eliminated (from 1 subject), the elimination half life is also significantly different between groups and inversely correlate with gestational age. However, volume of distribution is not significantly different between groups and no correlation with gestational age. In conclusion, pharmacokinetic parameters in Thai premature neonates are similar to data previously published from other countries.