Integration of Time-Varying Pharmacometric Modeling With Cox Regression for Time-to-Event Analysis in NONMEM

Area of Science:

• Pharmacometrics
• Statistical Modeling
• Drug Development

Background:

• Traditional time-to-event (TTE) analyses often use static drug exposure, ignoring real-world concentration variability.
• This limitation can lead to inaccurate pharmacokinetic (PK) or pharmacodynamic (PD) assessments for TTE endpoints.

Purpose of the Study:

• To develop and evaluate a methodology integrating time-varying pharmacometric models with Cox regression for TTE endpoints.
• To compare the performance of this semi-parametric approach against traditional parametric methods.

Main Methods:

• A simulation study was conducted using NONMEM software, incorporating a one-compartment PK model and a bathtub hazard function.
• The proposed semi-parametric method estimated parameters using partial likelihood via first-order approximation.
• Models were assessed using both static and time-varying exposure metrics.

Main Results:

• The semi-parametric method provided robust parameter estimates, especially when the baseline hazard was unknown.
• It demonstrated superior performance compared to parametric methods with incorrect baseline hazard assumptions.
• Utilizing time-varying exposure metrics significantly outperformed static metrics.

Conclusions:

• The proposed methodology effectively integrates time-varying PK/PD effects into TTE endpoint analysis.
• This flexible framework in NONMEM enhances model-informed decision-making for drug development, particularly for TTE outcomes.