Sunny Chapel, PhD, Jeffrey S. Barrett, PhD, Marc Pfister, MD
Aventis Pharmaceuticals, Bridgewater, NJ, USA
Objective: To explore differences in estimation methods to quantify food effect (FE) on absorption rate constant (Ka) between compartmental population modeling (CM) and noncompartmental methods (NCM). To test effects of model misspecification (fitting one-compartment model to two-compartment PK data), approximation methods (FO vs. FOCE interaction), fixed inter-individual variability of Ka, sampling schemes, number of patients, etc on the estimation performance.
Methods: One thousand six hundred FE 2-way cross-over trials were simulated using one-compartmental model with first order absorption. Sixteen trial designs were tested: 4 different sampling schemes (t-last 8, 12, 16, and 24h post-dose) and 4 different numbers of subjects (12, 18, 24, and 32 subjects/trial). FE on Ka (0 to 50%), Ka (1 to 6 times elimination constant) and interindividual variability on Ka (0 to 100%) were randomly assigned from uniform distributions. CM estimates for FE on Ka were compared with NCM. Prediction error (PE%) was calculated as 100(Estimated FE–True FE)/True FE.
Results: Regardless of trial design and approximation methods, CM resulted in |PE| >20% in less than 5% of trials. Model misspecification resulted in overestimated FE with PE >20% in more than 40% of trials. In contrast, NC Cmax ratios underestimated FE with PE <-20% in more than 90% of trials. With CM PE was not sensitive to trial design while NC tended to perform better with increased numbers of samples and subjects.
Conclusion: Compartmental population modeling is preferred to estimate the magnitude of FE on Ka. Model misspecification may significantly affect quantification of FE on Ka. The magnitude of FE on Ka estimated with NC Cmax ratios may be misleading. These results suggest that food effect labeling guidance may be inaccurate under certain conditions. These results must also be re-evaluated based on proposed exposure metrics, which may be more forgiving to noncompartmental methods.
Reference: PAGE 12 (2003) Abstr 452 [www.page-meeting.org/?abstract=452]
Poster: poster