Amira Ghoneim, Felix Boakye-Agyeman, Nathalie H Gosselin, Samer Mouksassi
1 Certara (Cairo, Egypt)
Overview: NONMEM has long been the industry standard for nonlinear mixed-effects (NLME) modeling, yet alternative platforms such as Phoenix NLME (QRPEM) are increasingly used when models show instability or fail to converge. During development of a sequential population PKPD model for an FDA submission, an existing population PK/PD model developed in NONMEM became unstable after the addition of Phase 3 data, preventing successful parameter estimation. Phoenix NLME was evaluated as an alternative platform for estimating the population PK/PD model. A benchmarking of the performance of the two platforms was conducted.
Methodology: A population PK/PD model previously developed in NONMEM 7.5.(FOCEI method) was updated with Phase 3 data using sequential approach. Due to convergence issues, final parameter estimates and individual post hoc values of PK model were imported into Phoenix NLME to develop the PK/PD model. Covariate effects were evaluated on PD parameters using a structured SCM procedure. Final PK and PK/PD models were simulated in Phoenix NLME to generate median profiles and 90% prediction intervals.
Results: NONMEM encountered convergence issues and numerical instability when the Phase 3 study was incorporated into the PK/PD model. In contrast, Phoenix NLME successfully estimated the full sequential model, producing stable and interpretable parameters. Phoenix NLME results were consistent with NONMEM estimates obtained prior to the addition of the Phase 3 data, demonstrating agreement between platforms. The SCM procedure identified relevant predictors for the PK/PD parameters, and simulated predictions showed good alignment with observed data.
Conclusion: Phoenix NLME provided a robust alternative when NONMEM was unable to complete the PKPD analysis. The QRPEM algorithm demonstrated improved numerical stability under challenging data conditions without compromising parameter interpretability. This case supports Phoenix NLME as a valuable tool for complex or computationally demanding NLME analyses, particularly in regulatory workflows where reliable model convergence is essential.
Reference: PAGE 34 (2026) Abstr 12026 [www.page-meeting.org/?abstract=12026]
Poster: Methodology - Estimation Methods