Using advanced computational tools has been a game changer for Teva api and our customers. This methodology has saved time, helped us achieve our Right First Time goals for scale-up and technology transfer, and optimized process development and production. Over the last few years, we have gradually increased the use of computational tools globally across Teva api with great success. Our customers benefit as we are able to sustain production and fill customer orders more quickly, in a more educated way, and with a higher quality and robustly manufactured products.
Building expertise to get the best results
Teva api uses several computational approaches and the best software packages on the market, including JMP by SAS for early stages of development, calibrated model-based process simulations using DynoChem, Visimix and CHEMCAD for comprehensive analysis and understanding crossing all stages (including technology transfers), and MATLAB by Mathworks/JMP for data analysis in multiple-batch production.
In order to get the right results, people need the right training. Instead of having one or two computational tools experts on staff, Teva api R&D decided to share knowledge and build expertise across our entire organization. We developed our own training program and trained local scientists and engineers at our R&D and manufacturing sites around the world. And, we created SME (Subject Matter Expert) groups, who serve as global support teams for all of Teva api users. Global SMEs are available to help colleagues “next-door” at their own local R&D or manufacturing sites, and may be called on to advise colleagues anywhere in the world. Each global SME group holds periodic meetings to share knowledge and experiences, discuss ongoing challenges, and enhance our common expertise – as well as lead together the group’s field according to Teva api’s vision.
Three computational approaches that optimize the product lifecycle and enhance expertise
Teva api uses three computational approaches during the API product lifecycle: Design of Experiments, Process Simulation, and Data Analysis.
1) Design of Experiments. Design of Experiments (DOE) is based on statistical analysis and is used mostly in the early stages of R&D in order to perform screening and conduct preliminary optimization of a new chemical process. It is supported by our Global DOE SME Group in Teva api R&D. A big advantage to using DOE is that it saves substantial development time and provides a better understanding of the process. We are able to extract maximum information from a minimal number of experiments, screen for significant process parameters and perform some optimization. It mainly allows us to bypass traditional trial and error methods to determine what is affecting a chemical reaction or other process through a well-planned set of experiments laying out our entire process design space.
When done properly, a project that has gone through DOE in development has a higher probability of achieving Right First Time goals in production, enabling Teva api to efficiently deliver a finished product that will help it to meet a customer’s exact specifications and deadlines. In one instance, DOE analysis led us to a more efficient process, enabling us to raise the yield of API from raw material by 15 percent.
2) Process Simulation. Mechanistic model-based process simulations can be used throughout the product lifecycle from process development to production. It is supported by the Power User Community: Process Simulation Global Group in Teva api R&D. Process simulations are scalable and are ideal for transfers from R&D to production. Teva api uses them for many processes including: chemical reaction (kinetics), solvent swap distillation, mixing, crystallization, filtering, heat transfer, and more. All are supported by reliable physical material property databases, as well as thermodynamic data for properties such as solubility and vapor–liquid equilibrium. In some processes, these simulations are performed as process feasibility trials before even getting into the lab in R&D, thus saving precious development time. Process simulation tools are supported with model fitting, optimization and design space, allowing for scalable calibrated models that are used during scale-up and in technology transfers between countries or production sites. They are also used in production when adapting process capacity, cycle time, or moving a process between vessels in the plant. Many times, these simulations serve as a decision making tool, determining the best vessel to transfer a process to, the optimal value of a process parameter (e.g., reagent addition time), etc.
3) Data Analysis. With data analysis, we can program codes in order to process production data, visualize, analyze, and then even create graphical user interfaces to export these analyses for further implementation by other users. Big data analysis is a new and evolving field, providing new statistical and other computational methods for utilizing large data sets accumulated in the new digital world we are now living in. We are looking into adopting selected methodologies from big data analysis to see how it can help meet both Teva api’s and our customers’ needs.
During production, data analysis plays a key role in everyday troubleshooting. When production problems occur and there’s an information gap, production data analysis helps us find the root cause, thus quickly focus on resolving it to stay on track so we can deliver finished product to customers as planned.
Case studies: Computational tools deliver proven results
There are many examples where computational tools have yielded tangible results for our customers and for Teva api. Here are two instances where simulations drived optimization:
Developing solvent swap distillation. For any mixture of solvents, we can use process simulation to develop a solvent swap distillation process as needed. In the early stages of development, the chemist starts with VLLE (Vapor-Liquid-Liquid Equilibrium) phase diagrams, setting the process pressure and temperature. Using the process simulation, they conduct a feasibility check – with no experiments necessary. The simulation will calculate whether the volume of fresh solvent to be added is reasonable. If the volume is too high, we save time by not proceeding to next steps in the lab. But if the volume works, we already have a scalable model that can be used from chemical R&D, to the pilot, to scale-up and technology transfer, and finally to production. The engineer just needs to update the system heat transfer values and system characteristics.
By using process simulation for solvent swap distillation, Teva api saves a substantial amount of work per case in R&D and better meets Right First Time production requirements. It also ensures that we follow a green process, part of Teva api’s commitment to manufacture products with a minimal impact on the environment, in this case by reducing the amount of organic solvents we use in our manufacturing processes. Our customers benefits by getting a cost-effective, quality product, helping them to meet their manufacturing timeline.
Continuous flow chemistry simulation. Our established methodology for flow chemistry is to first fit the reaction kinetics to experimental data (which we usually already have from previous work), then apply it to a predicting flow-chemistry system model. This allows for optimal system and process design for the first tests in the lab with that flow chemistry system, as well as serving as a feasibility trial to determine if it is worth proceeding to experimental trials with it at all. Case after case, we find these model predictions to be accurate and reliable, enabling us to promote this important new technology within a global organization without having to send people and materials to our central lab to conduct testing in a specific system. This is because this method enables us to first simulate the process on our computers with well-calibrated and validated simulations. Starting this venture, we learned that our preliminary “back of the envelope” calculations would have actually misled us. The simulation results corrected us and led us to configure the system entirely differently. In addition, we are proud to develop unique model-based simulations that are tailor-made for the new technology systems we evaluate. Following experimental validation, we share them with our global power-user community for them to implement in their R&D or manufacturing site.
Last, but definitely not least, the use of these models allows for better understanding of the process from a safety perspective through a valid simulation of the process’ dynamic response to its parameters and possible interferences. As safety always comes first, application of flow chemistry technology in general, and specifically using advanced models allowing for comprehensive process understanding, is of a great value.
Teva api’s strength lies in its experts
While the API industry is advancing with implementing new technologies that can be supported through various process models, my favorite model of all is the human model. Teva api’s strength comes from the expertise of our people and our multi-disciplinary collaboration, and the fact that we have broadly implemented our computational tools across the organization demonstrates that. Teva api R&D, engineering, and production teams are committed to sharing knowledge and experience to save time and effort across multiple sites and countries. Through conducting global training in-house and by building a team of experts with senior academic and industrial education, we bring industry-leading expertise to these processes. In Teva api’s culture of continuous improvement, we are always working on new techniques and refining our processes to deliver the best products and the best service to our customers around the world.
