Category: Manufacturing and Bioprocessing
Purpose: The viscosity of high concentration protein solutions can lead to a range of challenges in drug product manufacturing and administration. Accurately modeling the viscosity of biologics solutions in response to changes in the formulation and surrounding environment is of significant interest and remains a challenge. In this poster, we show a practical method of modeling the viscosity of a therapeutic solution in response to changes in both temperature and protein concentration. By better modeling changes in viscosity over the range of process-relevant concentrations and temperatures, viscosity related challenges in the drug manufacturing process, such as fill and finish activities, can be better anticipated and mitigated. In particular, time-pressure filling technology can be very sensitive to temperature excursions during the filling process. Accurately modeling the impact on viscosity of these temperature variations can improve the fill volume accuracy.
Methods: For each of the molecules in this study (including 2 IgG1s, 1 IgG2/4, and 1 trispecific antibody), the viscosities of 8 concentrations were measured ranging from 0 mg/mL to approximately 200 mg/mL in different formulations (11 total). For each concentration, the viscosity was measured at 7 different temperatures, ranging from 5 to 35°C. Viscosity measurements were made with a RheoSense Initium instrument, making sure that the samples do not exhibit any non-Newtonian effects such as shear thinning. This data was then fit to our model by first finding the activation energy, Ea, from an Arrhenious model of temperature dependence at each concentration. The Ea data as a function of concentration is then fit to a second order polynomial to provide for the thermal dependence on concentration in our model. Next, for a specific temperature, we fit the viscosity vs. concentration data to a Ross-Minton equation model. The best-fit parameters from this are then used to initialize a full non-linear regression fit of the data including both temperature and concentration dependence.
Results: Our model has two main components, the first being the concentration-viscosity term. We found experimentally that all the monoclonal antibodies used for this study were well described by the Ross-Minton model. The second component of our model is the Arrhenius temperature dependence of viscosity including both the standard exponential Arrhenius relationship as well as the concentration dependence of the activation energy, Ea. We found that with the addition of this concentration dependence to Ea, the Arrhenius model is able to accurately capture the temperature dependent changes to viscosity. Our model is able to accurate capture both the temperature and concentration dependence of viscosity for all of the molecules in this study with errors in predicted vs. actual viscosity being the same magnitude as the experimental uncertainty.
Conclusion: We have demonstrated that viscosity of therapeutic protein solutions can be successfully modeled through the combination of a Ross-Minton concentration dependent model with Arrhenius temperature dependence. This model can be used over a wide range of protein concentration, solution viscosity, and process-relevant temperature ranges. We have shown that this approach can be applied to different types of IgGs as well as a variety of typical therapeutic formulations. Overall, the accuracy of the model is acceptable for practical use and is similar in scale to the uncertainty inherent in experimental methods. Overall, the modeling done in this work helps to provide insight into the fill and finish process and how the rheology is affected by changes in environmental conditions. Knowing the sensitivity of the viscosity to changes in both concentration and temperature fluctuations allows for better understanding of the requirements and controls needed in the fill and finish process. In addition to this practical result, our data shows that solution viscosity and activation energy have a non-trivial correlation. This is potentially indicative of underlying mechanics connecting the two and warrants further study in future work.
Walter Schwenger– Framingham, Massachusetts
Walter Schwenger– Framingham, Massachusetts
Charlotte Pellet– Vitry sur seine, Ile-de-France, France
Delphine Attonaty– Neuilly-sur-Seine, Ile-de-France, France
Jean-rene Authelin– Vitry sur seine, Ile-de-France, France