PBPK-based exposure–response modeling of enzalutamide and its active metabolite N-desmethyl enzalutamide for the prediction of survival outcomes in patients with prostate cancer

Technical Summary: PBPK-based Exposure–Response Modeling of Enzalutamide for Survival Outcome Prediction

Problem Statement
Enzalutamide, a second-generation androgen receptor (AR) inhibitor, is the standard of care for metastatic castration-resistant prostate cancer (mCRPC) and metastatic hormone-sensitive prostate cancer (mHSPC), with an approved dosing regimen of 160 mg once daily. While effective, the drug is associated with treatment-related adverse events (e.g., fatigue, CNS toxicity) that often necessitate dose reduction or discontinuation. Although clinical observations suggest that reduced doses (e.g., 120 mg) may retain therapeutic efficacy, the pharmacological basis for such dose optimization remains insufficiently characterized. Specifically, it is unclear whether the additional systemic exposure provided by the standard 160-mg dose yields a clinically meaningful improvement in survival outcomes compared to reduced doses, or if it primarily increases toxicity risk. Existing physiologically based pharmacokinetic (PBPK) models have focused on drug-drug interactions and pharmacokinetic characterization but have not been sufficiently applied to quantitatively evaluate alternative dosing strategies in the context of long-term survival endpoints.

Methodology
The study developed an integrated modeling framework combining whole-body PBPK modeling with parametric exposure–response (E–R) survival analysis.

1. PBPK Model Development:

   • A whole-body PBPK model for enzalutamide and its active metabolite, N-desmethyl enzalutamide, was constructed using PK-Sim®.
   • The model incorporated physicochemical properties, ADME data, and clinical pharmacokinetic (PK) profiles from healthy volunteers and prostate cancer patients.
   • Metabolism was parameterized based on CYP2C8 (major pathway) and CYP3A4 pathways, with intrinsic clearance values optimized against clinical data.
   • The model was validated by comparing predicted vs. observed $C_{max}$ and $AUC_{last}$, calculating geometric mean fold errors (GMFE) and mean relative deviations (MRD).

2. Virtual Population Simulations:

   • Simulations were performed for 200 virtual individuals per dosing regimen (120 mg and 160 mg once daily) to characterize steady-state trough concentrations ($C_{trough,ss}$).
   • Pharmacological target attainment was assessed against a benchmark of 5 µg/mL (associated with near-maximal AR occupancy).
   • An exploratory safety benchmark of 16.1 µg/mL (mean $C_{trough}$ in patients requiring dose reduction due to Grade 3 toxicity) was used to contextualize exposure distributions.

3. Exposure–Response (E–R) Modeling:

   • Individual patient data (IPD) for progression-free survival (PFS) and overall survival (OS) were reconstructed from published Kaplan–Meier curves using the IPDfromKM algorithm.
   • Parametric survival models (Weibull for PFS; log-normal for OS) were developed using NONMEM.
   • The primary exposure metric was the PBPK-predicted total active-moiety $C_{trough,ss}$ (sum of parent and metabolite concentrations, assuming equal potency).
   • Models evaluated the effects of categorical dose groups (120 mg vs. 160 mg) and continuous exposure ($C_{trough,ss}$) on survival hazards.

Key Results

• PBPK Performance: The final model demonstrated acceptable predictive performance. GMFEs for $C_{max}$ and $AUC_{last}$ ranged from 1.16 to 1.35, and MRDs for concentration-time profiles were 1.39–1.41. Over 89% of predicted concentrations fell within two-fold of observed values.
• Target Attainment: Virtual simulations indicated that 77.0% of individuals on the 120-mg regimen and 85.5% on the 160-mg regimen achieved the pharmacological target of $C_{trough,ss} \ge 5$ µg/mL.
• Exposure Distribution: The proportion of individuals exceeding the exploratory toxicity reference of 16.1 µg/mL was 14.5% for the 120-mg group versus 28.5% for the 160-mg group.
• Survival Outcomes:
  • PFS: The model-predicted median PFS was 13.7 months (95% simulation interval: 8.7–20.0) for 120 mg and 18.0 months (11.0–25.0) for 160 mg.
  • OS: The predicted median OS was 32.0 months (22.0–50.0) for 120 mg and 34.0 months (32.0–36.0) for 160 mg.
  • While point estimates favored the 160-mg dose, the simulation intervals overlapped substantially between the two active-dose groups. The exposure-effect parameters had wide confidence intervals that included zero, indicating uncertainty in the magnitude of the independent exposure-survival relationship.

Significance and Claims
The authors claim that this study provides a model-informed pharmacological rationale for the prospective evaluation of reduced-dose enzalutamide (120 mg) as a dose-optimization strategy. The integrated framework suggests that the incremental systemic exposure achieved with the standard 160-mg dose does not translate proportionally into additional modeled survival benefit, particularly given the substantial overlap in predicted survival intervals. Conversely, the higher dose is associated with a greater proportion of patients reaching exposure levels linked to toxicity.

The paper explicitly states that these findings do not establish therapeutic equivalence or non-inferiority of the 120-mg dose, nor do they support replacing the current 160-mg standard of care. Instead, the work serves as a hypothesis-generating framework aligned with FDA Project Optimus guidance, supporting the need for prospective clinical trials to determine if a reduced dose can achieve a more favorable benefit-risk balance in selected patients. The authors acknowledge limitations, including the lack of individually matched PK and survival data and the assumption of equal potency between the parent drug and metabolite, emphasizing that the results are exploratory and require validation.