(M1030-11-64) Thaw Time Prediction for Partially-Filled Bottles of Antibody Solution in Varying Environments – Predictive Statistical Model on CFD-Generated Datasets

Purpose: The purpose of this research is to establish predictive equations for thaw time as a function of container fill volume and environmental conditions. Freezing/thawing (F/T) is an essential pharmaceutical process to ensure product quality and stability. The dynamics of F/T processes are contingent on several factors, such as thermodynamic properties of the product, fill volume, environmental conditions (frozen equilibrium and room temperature, symmetric/asymmetric airflow), etc. These parameters affect both thawing time and rate, which directly impact the formation of cryo-gradients, physical interactions at ice-liquid-air-interfaces, and thermal effects on macromolecular structures (e.g. Tg’), thus resulting in direct impacts on Critical Quality Attributes (CQAs). Hence, it is essential to understand how varying these environmental parameters across the full range of manufacturing conditions results in distinct thaw times and rates. In this study, we applied our novel computational fluid dynamic (CFD) model to assess the impact of partial fills on thaw behavior for a range of different environmental conditions. The resulting dataset was then statistically analyzed to yield operator-friendly equations predicting the time required for a given bottle to thaw.

Methods: Range of conditions: During clinical and commercial manufacturing, products are routinely frozen in 5L bottles (targeting 3.5L, ±10%) at either -80℃ or -30℃, then thawed under ambient conditions ranging from 15℃ to 25℃. We investigated bottle fill volumes from 20% to 150% of target, across this full range of environmental conditions. Computational model: The CFD model was built in-house using Ansys-Fluent to study the fundamental dynamics of F/T cycles. This model considers the thermal properties of a formulation (e.g. thermal conductivity, specific heat, latent heat etc.) and the environmental conditions impacting F/T rates (e.g. operating temperature, active/passive airflow, loading configuration, etc.) to simulate the processes across a wide range of scales and geometries. Dataset generation and analysis: The current study was conducted computationally using the CFD model to simulate thaw times for 32 different cases, spanning the aforementioned ranges. This enabled data to be generated faster than would be possible experimentally, while requiring fewer resources. Select conditions were then replicated in the lab for data validation. The computationally generated dataset was then statistically evaluated with an ANCOVA (analysis of co-variance) assessment to identify the significant parameters impacting thaw times. Finally, a quadratic regression model was built to generate predictive equations.

Results: The CFD model successfully replicated F/T profiles for different fill volumes across a range of manufacturing conditions. Continuous model validation efforts are currently underway with supporting experimental data. Although the model is capable of accounting for the thermal properties of specific products, we began building the predictive statistical model for thaw times from the most simplified set of conditions – assessing water in a 5L bottle. Dependance of thaw time on fill volume and environmental conditions: Evaluation of the resulting dataset indicated that total thaw time was predominantly governed by room temperature, with a quadratic relationship observed with fill volume, as shown in Figure 1. Differences in thaw times for varying fills: It is critical to understand how conditions in the lab, used to establish Prove-Acceptable-Ranges (PARs), compare to Normal-Operating-Ranges (NORs) encountered during manufacturing. Figure 2 shows the differences in thaw times between bottles filled to-target and those filled higher/lower, across common manufacturing conditions. This can be used to determine precisely how low the fill volume can be without exceeding the allowable hold time in a cryo-concentrated state, thereby rapidly enabling data-driven decisions on the acceptance/rejection of partially filled bottles for clinical and commercial campaigns. Predictive Equations: The ANCOVA analysis yielded two equations (Shown in Figure 3) for thaw time as a function of (1) both environmental conditions and fill volume, and (2) only fill volume for a given set of environmental conditions. While the former (Eqn. 1) is applicable in the absence of any empirical data, the latter (Eqn. 2) requires at least one supporting empirical data-point, and is therefore more suitable for late-stage applications in well-characterized environments. Both equations are operator-friendly and enable prediction of the time required for a given bottle to thaw.

Conclusion: The impact of fill volume and environmental conditions on thaw time was successfully studied resulting in two predictive equations. This approach can be applied to specific programs by inputting the relevant thermal properties (specific heat, latent heat, thermal conductivity etc.) and specific container geometries, environmental conditions, etc. to the aforementioned CFD model, in order to generate far more extensive and comprehensive data sets for statistical analysis than could ever be supported by experimental or manufacturing data alone.

Acknowledgements: Andrew Chiu, Daniela Marcano-James, Jordan Lu, Jeremiah Castro, Parvati Abdulpurkar, Douglas Kamen, Einer Santos Boulat, Jacob Weber, Emmanuel Jorge, Paula Lustgarten, Nino Gagliardi, Ross Kenyon


Figure 1: Relationship of thaw time with fill volumes for different environmental conditions


Figure 2: Differences in thaw times between bottles filled to-target and those filled higher or lower, across common manufacturing conditions


Figure 3: Predictive equations for thaw time as a function of both environmental conditions and fill volume (1), and only fill volume for a given set of environmental conditions (2)