How to handle population pharmacokinetics modeling in the context of a regulatory submission for a compound already well-characterized in other settings? A case-study with nivolumab in urothelial cancer.
Mélanie Wilbaux (1), Yue Zhao (2), Anna Kondic (2), Lora Hamuro (2), Amit Roy (3)
(1) Pumas-AI, inc., France, (2) BMS, USA, (3) Pumas-AI, inc., USA
Introduction: Population pharmacokinetic (PopPK) modeling is routinely used to characterize the PK of drugs over the life-cycle of its clinical development, including for indications subsequent to the initially approved indication. Much of the focus of the characterization of PK for the initial indication is on the covariate effect on PK, which is now an integral part of the overall clinical pharmacology profiling. In addition, the PopPK analysis enables characterization of exposure-response (ER) of biomarkers, efficacy, and safety in the target patient population to inform dose selection and justification. Characterization of the PK for subsequent indications facilitates the bridging of safety data from previous indications to the new indication and enables characterization of ER relationships to support dose selection and justification for the new indication.
When extending the use of a compound to new settings (e.g. indication, population), decisions regarding dosing strategies are informed by clinical trial data and PK/pharmacodynamic (PD) considerations. In such scenarios, questions emerge regarding whether the previous PopPK model has been adequately characterized and validated to accurately represent the new clinical data and be utilized for generating individual exposure.
Nivolumab, a PD-1 inhibitor, serves as an illustrative example due to its prior approvals for multiple indications (e.g. non-small cell lung cancer (NSCLC)), both as monotherapy and in combination with other treatments. The PK of nivolumab has been extensively characterized [1]. Here, we present a case study utilizing new Phase 3 data evaluating nivolumab in combination with chemotherapy for untreated unresectable or metastatic urothelial carcinoma (UC), aiming to characterize its PK and derive exposure for ER analyses and support regulatory submission.
Objectives: To explore different methodologies of handling PopPK modeling for a compound already well-characterized in other settings, utilizing nivolumab as a case-study.
Methods: The PK of nivolumab is well-described by a 2-compartment model with time-varying clearance (sigmoidal-Emax). Previous characterizations indicate similar clearance across different tumor types and treatment combinations. Therefore, the PK of nivolumab plus chemotherapy in UC was expected to be comparable to the previous indications.
An external validation approach using a previously developed PopPK model was used to obtain empirical Bayesian estimates of individual PK parameters, given the previously estimated population parameters (MAXEVAL=0 in NONMEM) and the PK data, covariates, and dose history from all subjects. The adequacy of the model was assessed using prediction-corrected visual predictive checks (pcVPCs). In case of bias observed in pcVPC, the model was re-estimated using pooled data from historical and new data and refined following two steps. First, the Base Model consisted of re-estimation of all the parameters. Then, the Full Model was developed by incorporating the effects of combination with chemotherapy on time-varying clearance (Emax) and baseline serum albumin on baseline clearance (CL0), both of which were found to have a significant effect in other previously developed PopPK model.
Results: The nivolumab PopPK analysis dataset included 6518 concentration values from 1355 subjects with melanoma, NSCLC, or UC who received nivolumab monotherapy or combination therapy with chemotherapy from several clinical studies. External validation approach based on pcVPC showed a strong bias toward underprediction for the new data. The model showed significant improvement after re-estimating model parameters and incorporating additional covariate effects. Nivolumab CL0 was 23% higher at the 5th percentile of baseline serum albumin compared to the reference value. The extent of change in nivolumab CL over time (Emax) in subjects treated with nivolumab in combination with chemotherapy was 6.1 % lower than in those treated with nivolumab a monotherapy.
Conclusions: This work provides a comprehensive framework for PopPK modeling of a well-known compound in new therapeutic contexts, using nivolumab as a case study. Initially, prioritizing external evaluation methods is recommended to avoid unnecessary redundant model development, a particularly valuable approach when faced with tight deadlines. Model refinement is then undertaken only in the presence of significant bias, indicating that potential other factors may affect PK variability between studies and patient populations.
References:
[1] Zhang J et al. (2019). Population Pharmacokinetics of Nivolumab in Combination With Ipilimumab in Patients With Advanced Malignancies.
