Corinna Maier, Sven Mensing, Insa Winzenborg
AbbVie Deutschland GmbH & Co KG
Objectives:
The exploration of the relationship between drug exposure and clinical efficacy and safety response endpoints is vital for informed dose selection and dose confirmation in drug development. Logistic regression approaches are commonly used due to ease of understanding, few assumptions and efficient implementation in contrast to more elaborate mechanistic pharmacodynamic (PD) models that consider the longitudinal nature of the underlying data. While this approach of using one-point summaries for PK as well as efficacy and safety is appealing in face of fast-moving drug development timelines, it may overlook important factors and introduce analysis bias. Potential biases introduced with the exposure metric average concentration up to the event (CavgTE), a commonly applied concept in oncology, have already been explored previously1,2. This study aims to contribute to this discussion by exploring the direction of potential biases for different dosing designs, including step-up or ramp-up designs, dosing holidays and dose reductions.
Methods:
An interactive simulation framework was built to explore different settings of interest. A two-compartment model with linear absorption was used to describe the pharmacokinetics of an orally administered drug, which was linked to a Friberg-type pharmacodynamic model3 to simulate the effect on blood cell counts, assuming a direct drug effect on circulating cells as well as a drug effect on proliferating cells in the bone marrow which has a delayed effect on blood cell levels. Inter-individual variability was included on clearance, central volume of distribution of the drug as well as on baseline cell counts and both drug effects. Two exposure metrics were considered. CavgTE is calculated by averaging the simulated pharmacokinetic profile up to the first occurrence of the event of interest in case an event occurs, or up to the end of treatment if no event occurs. The second considered metric is steady-state AUC (AUCss), calculated as steady state dose divided by apparent clearance. The explored safety endpoint Grade 3+ adverse event (AE) was derived from simulated cell counts below a certain threshold including a proportional error.
Results:
Simulations of a bi-weekly ramp-up scheme starting with 1/20th of the target dose and increasing doses over time showed that CavgTE can lead to a negative exposure-response relationship, despite having a simulated positive drug effect, i.e. a drug effect that leads to a greater reduction of blood cells with higher doses. On the other hand, using steady-state AUC as the exposure metric leads to a meaningful exposure-response relationship, describing a higher rate of Grade 3+ AE with higher exposure. Furthermore, a 14 days on, 14 days off regimen was explored, allowing for the recovery of cells during the off period. In this scenario, an apparent steeper exposure-response relationship was found with CavgTE compared to QD dosing for 28 days, as early events are linked to higher CavgTE exposure and subjects without event are evaluated at the end of the cycle associated with lower CavgTE exposure. Similarly, dose reductions and interruptions as frequently occur in oncology trials lead to steeper ER relationships with CavgTE as the corresponding regular full dosing scenario.
Conclusions:
This simulation study demonstrates that significant bias can be introduced in endpoint exposure-response models by calculating average concentration up to event as exposure metric. A priori simulations using mechanistic models can assist in identifying potential biases and in defining appropriate exposure metrics before analyzing the data. If developing pharmacodynamic models at the stage of data availability is not deemed possible, e.g. due to short decision or submission timelines, these a-priori considerations can significantly enhance the quality of internal decision-making as well as regulatory submissions and prevent analysis bias.
Overall, this simulation study underscores the importance of carefully selecting aggregated exposure metrics and conducting a-priori simulations when exploring exposure-response relationships to inform dose selection and confirmation throughout the various stages of drug development.
References:
[1] Wiens MR, French JL, Rogers JA. (2023). Confounded exposure metrics. CPT Pharmacometrics Syst Pharmacol.; 00: 1-5. doi:10.1002/psp4.13074
[2] Largajolli A et al. (2023). Considerations when deriving time-averaged exposure for censored subjects for logistic regression exposure-response analyses. PAGE meeting url: https://www.page-meeting.org/default.asp?abstract=10425
[3] Friberg L et al. (2002). Model of chemotherapy-induced myelosuppression with parameter consistency across drugs. Journal of Clinical Oncology, 20(24), 4713–4721.
doi: 10.1200/JCO.2002.02.140
Disclosures:
All authors are employees of AbbVie and may hold AbbVie stock. This study was sponsored by AbbVie and AbbVie contributed to the study design, research, and interpretation of data, and the writing, reviewing, and approving of the publication.
Reference: PAGE 32 (2024) Abstr 10815 [www.page-meeting.org/?abstract=10815]
Poster: Methodology - Other topics