In this study the pharmacokinetics of the aminoglycoside amikacin was analyzed in 651 patients, and the parametric method IT2B was compared to the nonparametric method NPAG. Both methods are programs in the USC*PACK software. The total number of serum levels analyzed was 2641, ranging from 1-21 levels per patient, with an average of 4.06. A two-compartment model was first fitted using IT2B. Since a multi-center study was done, the overall essay error polynomial describing the SD of the intraindividual variability was first obtained. Second the median parameters describing the model was found with both the IT2B and the NPAG programs. The overall second order polynomial obtained with IT2B was SD= 0.3035 + 0.1174C1+0.00004676C2, where C is the serum concentration. Significant correlation was found between central compartment volume of distribution (VOL) and body weight, and between creatinine clearance (CCr) and the elimination rate constant (KEL). To reflect these covariate relationships, VOL and KEL were changed to VS1 and KS1, respectively. The median parameter values and their SD obtained with IT2B were: absorption rate constant (KA) = 1.44 ± 1.90, rate constant between central and peripheral compartment (KCP) = 0.0493 ± 0.0443, rate constant between peripheral and central compartment (KPC) = 0.00239 ± 0.00135, KS1 = 0.00225 ± 0.000836, VS1 = 0.335 ± 0.0915. The log-likelihood was 7257. The same values obtained with NPAG were: KA = 1.50 ± 1.48, KCP = 0.0745 ± 0.128, KPC = 0.133 ± 0.199, KS1 = 0.00302 ± 0.00145, VS1 = 0.297 ± 0.0982. The log-likelihood was 6482. The NPAG model was also implemented into the MMLQ program for validation on how well individual patients in the model would be predicted from the model. For all parameters, except the KA, the population CVs were smaller with IT2B, suggesting more precise population parameter estimates than with NPAG. However, since the log-likelihood was better with NPAG than with IT2B this indicates that NPAG was more able to perceive the true population diversity. The NPAG model obtained seemed to be well suited for planning initial amikacin dosage regimens with the MMLQ program. To obtain a better model of amikacin, a study design planning for more serum levels drawn per patient over a longer time period would be necessary.