(M1430-09-48) Combining Arrhenius Kinetics and Monte Carlo Simulation to Enhance Stability Prediction for Antibodies

Purpose: Ensuring the long-term stability of therapeutic antibody drugs is a crucial component of formulation development. Traditional methods for evaluating the long-term stability often require assessments lasting over a year. As a result, the use of Arrhenius kinetics-based modeling to predict long-term stability from accelerated stability data collected over a short period (usually 1-3 months) has garnered significant interest (1). This approach presents a valuable opportunity for application in formulation development and candidate developability assessment. Traditional Arrhenius kinetics-based modeling, however, lacks the ability to provide confidence intervals for its predictions, which hinders the assessment of worst-case scenarios and limits its practical utility in long-term stability assessment. Additionally, selecting appropriate kinetic models for predicting antibody aggregation remains challenging due to the complex underlying reaction mechanisms. This study aims to develop an Arrhenius kinetics-based modeling tool that can generate confidence intervals for its predictions using Monte Carlo simulation. In addition, this study evaluates the performance of two kinetic models in predicting antibody aggregation at long-term storage conditions and the utility of the Arrhenius kinetics-based modeling tool in candidate developability assessment.

Methods: The Arrhenius kinetics-based modeling tool was developed in JMP (JMP 17.0, JMP Statistical Discovery). Monte Carlo simulation was employed to incorporate data variability into the model and generate prediction confidence intervals. The confidence intervals were generated based on 10,000 times of Monte Carlo simulations. Stability data from stressed (32°C and 40°C/75% RH), accelerated (25°C/60% RH), and/or storage conditions (5°C) for up to 3 months were used. The tool's accuracy was validated using real-time storage stability results (2-8°C for up to 3 years). Quantification of high molecular weight (HMW) species, a critical quality attribute, was performed using size-exclusion Ultra-high Performance Liquid Chromatography (SE-UPLC).

Results: Arrhenius Kinetics-Based Modeling Tool Development and Evaluation - An Arrhenius kinetics-based modeling tool was developed in this study using JMP scripts. This modeling tool requires minimal user inputs and can generate a 95% confidence interval for its predictions based on the user input of assay variability. We evaluated the performance of Arrhenius modeling based on both zero-order kinetics and empirical non-linear kinetics in predicting HMW formation at the storage condition (2-8°C) across 13 antibodies and antibody-like molecules, including IgG1, IgG4, bispecific antibodies, and engineered antibodies. The Arrhenius modeling based on zero-order reaction kinetics demonstrated good accuracy in predicting the 1-year storage stability at 2-8°C. However, it significantly overestimated HMW formation for time points beyond 1 year. In contrast, the empirical non-linear kinetic model showed excellent accuracy in predicting HMW formation over a period of up to 3 years. These findings highlighted the utility of the empirical non-linear kinetic model in predicting antibody aggregation during long-term storage at 2-8°C. Application in Candidate Stability Risk Assessment - We further evaluated the application of the Arrhenius kinetics-based modeling tool in candidate risk assessment for three novel engineering antibodies. Stability data of the candidates at 25°C, 32°C and 40°C for up to 1 month were used to develop the Arrhenius models. Despite a significant increase in %HMW – surpassing the historically established low-risk threshold – after storing for 1 month at 40°C (Figure 1A), antibody 1 showed only a minimal increase in %HMW –total %HMW well below the specification – following 1 year of storage at 2-8°C (Figure 1B). The predicted increase in %HMW after 1 year at 2-8°C closely matched the observed increase (Figure 1B), demonstrating the utility of the modeling tool in predicting long-term stability based on short term accelerated data. Furthermore, the tool was capable of calculating the activation energy for HMW formation, offering insights into the high temperature dependency observed in HMW formation (Figure 1C). These results highlight the limitation of accelerated stability-based candidate risk assessment and underscore the advantages of the Arrhenius kinetics-based modeling in candidate risk assessment.

Conclusion: We have developed an Arrhenius kinetics-based modeling tool that generates a confidence interval for its prediction through Monte Carlo simulation. This feature aides in assessing worst-case scenarios, thereby enhancing the practical utility of Arrhenius modeling in long-term stability assessment compared to traditional methods that do not generate any confidence metrics. Designed as a JMP script-based tool, it streamlines the model development process, which traditionally requires multiple manual data manipulation steps. This study compared two different kinetic models and revealed the utility of an empirical non-linear kinetic model in predicting antibody aggregation at 2-8°C for up to 3 years. Zero-order kinetic models demonstrated good accuracy for predicting antibody aggregation for up to 1 year, but overestimated aggregation for timepoints longer than 1 year. Furthermore, this study showcased the utility of the modeling tool in candidate developability assessment. Compared to the accelerated stability-based approach, Arrhenius kinetics-based modeling tool provides a direct prediction of long-term stability and offers insights into the temperature dependence of degradation pathways.

References: 1. Drago Kuzman, Marko Bunc, Miha Ravnik, Fritz Reiter, Lan Zagar, Matjaz Boncina. Long-term stability predictions of therapeutic monoclonal antibodies in solution using Arrhenius based kinetics, Sci Report, (2021) 11: 20534.

Figure 1. Utilizing the stability modeling tool for candidate stability assessment