Comparison of Parametric (NONMEM®) and Non-Parametric (NPEM®) Methods for Population Pharmcokinetic Modeling of Bi-Model Populations
Purpose: Debate exists whether the most appropriate population analysis method is parametric (NONMEM®) or non-parametric (NPEM®), especially for data from a bi-modal population (i.e., poor/extensive metabolizers (PM and EM)). This simulation study compared the capability of NONMEM® and NPEM® to estimate PK parameters for a bi-modal population.
Methods: Concentration data were simulated using a one compartment PK model with first-order absorption and elimination. The PK model was fit to the simulated datasets using NONMEM® (first-order: FO and first-order conditional: FOCE) and NPEM®. The bias (PE%) and precision (|PE|%) of the individual predicted estimates of clearance (CL) and volume of distribution (Vc) were calculated. The log-transformed individual estimates of CL and Vc were tested for a statistical difference between the PM and EM subjects. A sign test was conducted to test for statistical differences in bias and precision of estimates for NONMEM® versus NPEM®.
Results: The model minimized successfully for all datasets and methods except for two datasets using NONMEM® FOCE. The predicted CL was statistically different for the PM and EM subjects. Estimates of Vc did not achieve statistical difference for all datasets and methods. All methods were able to predict CL with minimal bias (< ± 6%) and a high degree of precision (< 19%). On average the median PE% for Vc was -0.9%, -12%, and 5.5% for NONMEM® FO, FOCE, and NPEM®, respectively. The 75th percentile|PE|% of Vc on average was 28%, 38%, and 52% for NONMEM® FO, FOCE, and NPEM®, respectively. The estimates of CL for NONMEM® FOCE were statistically less biased than NPEM® and the estimates of CL and Vc for NONMEM® were statistically more precise than NPEM®.
Conclusion: NONMEM® and NPEM® adequately estimated the PK parameters for a bi-modal population. NONMEM® PK estimates were generally more precise than NPEM®.
American Association of Pharmaceutical Scientists (AAPS), Salt Lake City, Utah, October 2003
By L. Phillips, M. Vo, J. Hammel, J. Fiedler-Kelly, and E. Antal