NONMEM, a computer program that uses the method of extended least-squares analysis, has been advocated as a means of obtaining estimates of population pharmacokinetic parameters when only fragmentary information can be obtained from subjects. To assess the performance of this program, we compared NONMEM with traditional methods for the estimation of population pharmacokinetic parameters with data collected during a phase III clinical trial of alprazolam. NONMEM estimates of the population mean clearance and its coefficient of variation were identical to the estimates obtained with traditional pharmacokinetic techniques. Moreover, NONMEM estimates of these parameters remained stable even when a few as three data points were available per subject. NONMEM estimates of the mean volume of distribution and its coefficient of variation appear to be overestimated, apparently because of the sampling scheme used to generate data for the NONMEM analysis. Suggestions for the effective use of NONMEM in clinical trials, to maximize the benefits of this approach, are provided. Our results lend further support for the use of NONMEM to estimate population pharmacokinetic parameters of a drug from data generated during phase III clinical trials.
By Thaddeus H. Grasela, Antal EJ, Townsend RJ, Smith RB