Predicting Lyophilization Performance
Lyophilization, or the removal of water at low temperatures, is an effective means of preparing shelf-stable drug products with retained efficacy that otherwise cannot be formulated as sterile, ready-to-use aqueous solutions, which is often the case for protein-based therapies. To be successful, however, the appropriate conditions must be used. The formulation composition (e.g., buffer, bulking agent, stabilizer, surfactant, and other excipients) can affect the stability of the freeze-dried cake and the reconstituted product, as well as the processing temperature, which in turn affects the cycle time, and thus costs. The vial design and positioning of shelves in the lyophilizer are also important. Operating parameters, mainly temperature and pressure, are also crucial. Inexpensive methods for the rapid screening of formulations and for measuring product conditions throughout the lyophilizer in a non-destructive manner have not yet been developed. One area that has been progressing, however, is the development of effective process modeling tools and process design strategies that can help reduce the amount of trial and error needed to optimize lyophilization, according to Robert Sever, business development manager for life-sciences and laboratories at Praxair.
Modeling for processdevelopment and improvement
“We focused on the primary drying part of the cycle (when the frozen water is sublimed), because this step accounts for ~80% of the overall process time and consumes the largest amount of energy, both of which should ideally be minimized,” Koganti notes. Two different models were employed. The first was a simpler, steady-state model that can be solved in Excel. The second is a more complex, non-steady state model consisting of a collection of partial differential equations that is solved using commercially available finite element method (FEM) software for freeze-drying applications.
In the lyophilization process, a solution of the API and other excipients in water is filled in a partially stoppered vial. The solution is then frozen, and the water is sublimed by reducing the pressure and increasing the temperature. The key inputs, therefore, needed for any model of the freeze-drying process including the operating parameters (e.g., shelf temperature and chamber pressure) are the heat transfer efficiency of the vial and the resistance to moisture release in the vial. “The heat transfer efficiency of the vial indicates how much heat can be transferred from the shelf of the lyophilizer to the product via the vial, while the resistance to moisture indicates the level of resistance that water from the lower layers in the vial feels as it passes through the upper layers that have already dried,” Koganti explains.
The models were then applied to lyophilization processes in three settings: process development, improvement of existing processes, and troubleshooting problems in a manufacturing setting. At the product/process development stage, the use of these models for the freeze-drying process can be used to simulate the primary drying process and replace many of the actual, physical laboratory-scale runs that would typically be needed, which helps reduce the development time, according to Koganti.
For existing historical lyophilization cycles, in many cases the primary drying processes have not been fully optimized, and there is significant potential to reduce the cycle time, increase plant utilization, and thereby reduce costs. In this application, a model is built that simulates the current process and performance, and then the operating parameters are adjusted to determine the optimum conditions for reducing the cycle time without affecting the critical quality attributes of the product. “Here again, access to these models enables the reduction of the number of physical experiments that are required,” Koganti says. Finally, he notes that in the manufacturing setting, there are numerous problems that occur, and models can be used to relate the performance to first principles and then tailor the conditions to avert or avoid them. These models are also good for predicting the behavior of a process that is being transferred from one lyophilizer to another or one site to another.
Most importantly, the application of the models to all of these types of lyophilization processes has been successful. For Koganti, a successful prediction provided a drying time within 80% of the actual time observed and a temperature within two degrees of the actual temperature required. In the few cases where the model did not do a good job at predicting performance, the estimates for the heat transfer coefficient and/or the resistance to moisture transfer were typically inadequate. “When we were able to obtain better quality inputs, the models worked very well, and overall they enabled us to reduce the number of needed physical runs and increased the efficiency of product development, improvement, and trouble-shooting activities,” he asserts.
Modeling for effective scale-up
To address this problem, Sadineni and his colleagues developed a modeling approach to predict the commercial process parameters for the primary drying phase. Based on the use of experimental lab data and finite element modeling, they identified an efficient method for the scale-up of
“Our approach involves the use of a combination of computational and experimental data to predict commercial process parameters for the primary drying phase of lyophilization,” says Sadenini. Heat and mass transfer coefficients are determined based on manometric temperature measurement (MTM) experiment data and are used as inputs for the FEM-based software called PASSAGE, which computes primary drying time for various primary drying parameters, such as shelf temperature and chamber pressure. The PASSAGE software simulates freeze drying in vials and pans and can be used for both primary and secondary freeze-drying steps. The time for drying, along with temperature, vapor concentration, water pressure, and glass transition distributions can be calculated using the program.
“With this experimental/modeling approach, we are able to use the results from just five to six lab-scale runs to predict the optimum conditions under which we should perform the corresponding commercial-scale process. Because of the simple and minimalistic nature of this approach, it is less capital-intensive and can be executed with minimal use of expensive drug substance/active material,” Sadineni says.
Understanding excipient states
In addition, some excipients have the ability to form a glass or a crystal; typically the crystalline state is more stable, and the amorphous state is considered metastable. If the amorphous state is desired for protein stability, then crystallization of the lyoprotectant will leave the protein exposed to extremely harsh conditions during lyophilization, such as increased salt concentration, changes in pH, increased mobility, and favorable conditions for protein unfolding, which inevitably results in product degradation, according to Cullen. “Applying inappropriate process parameters to a biopharmaceutical formulation may result in having an excipient in an incorrect state as required by the product. Therefore, it is important to select the correct process parameters that render each excipient in its preferred state,” Cullen says.
Although the states that excipients take and their glass transition and/or eutectic temperatures are well known, when formulated, they exhibit thermal characteristics that are a compromise of each individual behavior. To predict this modified behavior, thermal characterization (e.g., modulated differential scanning calorimetry and freeze-drying microscopy) and X-ray diffraction analyses are used to determine the appropriate lyophilization process parameters that will render each excipient in its correct state for optimal stability and more efficient processing of a given formulation, according to Cullen. Other analytical techniques, such as Fourier transform infrared spectroscopy and size-exclusion chromatography provide information on protein structure and aggregation behavior.
Cullen and his colleagues are investigating excipient state formation to gain further understanding and knowledge of the different characteristics of Genzyme’s products. The results will be used to offer further technical support during manufacturing and technology transfers while also informing product technical lifecycle projects of process and product optimization.
Predicting the effect of water on stability
“There are several possible mechanisms for how water influences the chemical stability of amorphous freeze-dried products. Water as a plasticizer for increasing molecular mobility and thus chemical activity is the most commonly discussed option. However, we have found that if we look at a reaction similar to the deamidation of proteins, such as amide hydrolysis in mixed solvents, which can be influenced by water via polarity, apparent acidity, and other effects, we find that the relative change in the reaction rate for the reaction in solution is similar to that in freeze-dried formulations in the glassy state. Other observations that highlight the importance of media effects (at the expense of global molecular mobility) include: amide hydrolysis in the glassy state is under reaction (not diffusion) control; there is a strong correlation between the apparent acidity (expressed as the Hammett acidity function) and the degradation rate in a number of freeze-dried systems; and there is a lack of acceleration of the degradation rate in the rubber state as related to the glassy state in several systems. Therefore, global molecular mobility is likely not the predominant mechanism, and a solvent medium effect in which the water catalyzes the reaction, perhaps by interaction with the active site and lowering the activation barrier, is more likely,” Shalaev explains.
Unfortunately, there is currently no way to predict the appropriate residual water level for a given formulation, and thus physical screening of various formulations with different residual water contents is necessary, according to Shalaev. He adds that there are four ways to prepare samples with different water contents for such studies, and the choice of a particular method should depend on both practical considerations (e.g., the number of samples to be made) and a mechanistic understanding of the potential non-equivalency of the physical properties of the samples prepared using these different methods (e.g., the extent of relaxation in the glassy state, the apparent acidity, and the local structure). “Finally, it should be stressed again that knowledge of the phase composition of the freeze-dried cake, including the extent of crystallinity of the excipients, and monitoring the changes in the crystallinity during stability testing is important, because it can impact the local water content and thus the stability of the product,” Shalaev observes.
About the Author