Gaining Experience in Continuous Manufacturing

Continuous manufacturing (CM) of oral solid dosage drugs has benefits that include faster development, a smaller equipment footprint, greater flexibility in manufacturing scale, tighter process control (and, thus, improved quality), and the potential for real-time release. Contract development and manufacturing organization (CDMO) Hovione partnered with Vertex in 2016 to expand Hovione’s facility in East Windsor, NJ, with continuous manufacturing equipment to produce Vertex products. Pharmaceutical Technology spoke with CM team members at Hovione: Alexandra Adao, head of Quality Assurance Continuous Manufacturing; Sarang Oka, process development engineer; and Jose Luis Santos, head of Drug Product Continuous Manufacturing about their experiences with implementing CM.

Benefits and challenges

PharmTech: As CM processes for solid-dosage drugs are being commercialized, what benefits are being realized? What challenges are arising?

Santos (Hovione): CM is a tremendous opportunity to advance drugs to the market at a much faster pace than traditional batch manufacturing, and additionally enabling enhanced product quality based on enhanced process understanding and a richer data environment.

Investment is significant, not just in terms of the equipment itself, but also the software for integration, which should include a strong [process analytical technology] (PAT) framework; the facility to host the equipment; and all the utilities associated.

The teams to be put together may be much different from the standard teams that typically support batch solid dose manufacturing. Now the skill-set should comprise a strong background as well on process modeling, automation and control. PAT is now a mandatory role in the teams. The quality teams should either adjust themselves to embrace a new paradigm, or new focused teams should be put together.

Development is more difficult, and requires more sophisticated process understanding, including process modeling. Currently it’s probably not as rapid as batch development, given the steep learning curve for most companies. It is likely that development takes more time than batch for similar development phases, but it shows a clear potential to be much faster than batch in the full cycle of development of a drug, since development is conducted always in the same equipment and same scale.

Even if our experience is still limited, the numbers we have currently at Hovione in terms of equipment occupancy already surpass those from our typical batch processes. So, there is a significant opportunity to make much better use of the assets in continuous as compared to batch manufacturing.

PharmTech: What are some of the challenges for feeding materials into the system, and what are some of the best practices that have been implemented?

Oka (Hovione): Two challenges in feeding that plagued the technology in its early days continue to persist at their core, namely, feeding at very low flowrates and feeding of poorly flowing powders that exhibit substantial electrostatic adhesion. Although substantial progress has been made in how the challenges are managed, there has been very little transformative change in technology of powder feeding. Loss in weight (LIW) screw feeding continues to be the dominant feeding technology. For example, a few years ago, an academic group prototyped the use of a salt-shaker type feeder for feeding at very low flowrates. The technology showed only limited success, and thus, in absence of any transformative digression from LIW screw feeding, we continue to be constrained by physical limitations of screw feeders.

As previously mentioned, we have, however, made substantial progress in how we manage the challenge when feeding poorly flowing powders or feeding at low flowrates. The primary challenge with poorly flowing powders is inconsistent filling of the screw pitch. Vendors have experimented with various agitator and screw geometries to improve screw filling, and these have shown promising results. Some powders, such as certain grades of silica with very low bulk density and substantial electrostatic adhesion tendency, however, continue to pose a challenge. Another trick in the book is coprocessing the poorly flowing powder with a flow aid prior to introducing the powder to a feeder. For example, poorly flowing active ingredients are preblended with silica to improve their flow properties prior to being fed via screw feeders. In terms of excipients, progress has been made to design and develop excipients that are well suited for processing in continuous manufacturing lines and don’t suffer from the aforementioned challenges.

In the case of feeding at low flowrates, we are largely constrained by physical limitations of the hardware. There is a physical lower limit below which signal-to-noise ratios even for the most sophisticated load cells becomes too low for effective gravimetric control, and this handicaps our ability to feed below a certain flowrate. A management strategy has been to preblend (in batch) the ingredient with low weight fraction in the formulation with another ingredient, and then feed the preblend at the combined, higher feedrate.

Tech transfer and scale-up

PharmTech: What are some best practices for technical transfer or scale-up that are unique to or particularly important for CM processes?

Oka (Hovione): These are many advantages when developing products for continuous processes. Development is performed at scale, eliminating the need for performing rigorous scale-up studies as would be the case for batch processing. One can be very confident that experiments performed and results obtained on a standalone loss-in-weight feeder will translate well when an identical feeder is used to feed the same material in the integrated process train. Tech transfer is often performed between very similar (in size and other characteristics) or identical units. The challenge arises when development is performed on equipment that is different from commercial equipment. We still do not have correlations that help us transfer process between dissimilar pieces of equipment or process trains. Recommended practice for organizations who wish to build multiple lines is to develop identical lines, if there is desire to transfer products between lines.

Interestingly, since several unit operations in the batch paradigm are intrinsically continuous (such as roller compaction or tableting), practitioners have been able to rely on their current development practices built over decades of development knowledge. For example, the use of compaction models and compaction simulators to develop roller compaction or tableting processes in batch manufacturing have been applied to developing these unit operations for continuous processes. Practitioners have been able to directly transfer such best practices developed for the batch world and apply them to developing continuous processes. Of course, this is not possible in cases where the unit operation is different from a batch process, such as powder blending, and completely novel development frameworks have had to be built for these.

Process modeling

PharmTech: How are process models/digital twins being used in CM process development/process control? What are the next steps for the near future?

Oka (Hovione): Serious effort is being made in the development of digital twins for continuous manufacturing process trains and individual unit operations, with the underlying modeling framework ranging from discrete element method (DEM) based, population balance models, or other empirical and semi-empirical approaches. A very popular application of digital twins has been to perform in-silico residence time distribution (RTD) experiments. These experiments can be done either at the unit operation level or at the full process train level. Thorough characterization of the RTD enables one to track evolution of disturbance through the rig, which is a crucial element of today’s control schemes. The underlying models, if well calibrated, also enable the practitioner to ‘dial-in’ and test a desirable RTD, all performed from the comfort of one’s desk. Applications also include understanding the impact of material properties and process parameters on process performance and eventually product quality. Some digital twins even predict product performance, that is, the performance of the drug within the body.

It is also important to mention the drivers for these efforts. The primary driver continues to remain circumventing the use of valuable early stage API. However, practitioners of the technology are beginning to realize that most CM rigs have large turnaround times for cleaning. Introducing the material into the rig for experimentation and ‘dirtying’ the rig is associated with the cost of the follow-up cleaning. This limitation is especially important to us, as a CMO, where this translates to a substantial opportunity cost. Tools that help us minimize the use of the rig are thus incredibly important to us. These could be very sophisticated models or simple tools based on engineering correlations that predict the performance of the product in the rig.

Real-time release

PharmTech: Is real-time release happening? What roadblocks remain?

Adao (Hovione): Hovione has already introduced all the requirements to enable real-time release testing in the quality systems. There are some companies that have already successfully implemented it, and we are confident that we have in place all the conditions to also perform its implementation. 

There are some technically relevant aspects that enable the implementation of real-time release testing, namely the online and inline PAT capabilities and the design of automation control systems (e.g., equipment monitoring, material tracking). An appropriate control strategy that combines both capabilities is a key element for an adequate implementation of the real-time release in CM. At Hovione, the referred capabilities are already in place and system qualification was performed, taking also into consideration the requirements for real-time release testing.

An example of a potential challenge in the implementation of real-time release is dissolution testing. In order to enable real-time release, the use of model-based prediction of dissolution performance based on process parameters and critical quality attributes (CQAs) will be required, and consequently, an adequate model maintenance procedure needs to be in place. The model maintenance requires a parallel testing program to test the methods against the respective laboratory/regulatory methods. Therefore, in addition to the significant effort associated with the verification of the models during implementation, there is also an important effort associated with the model lifecycle.

As a CMO, we have been preparing ourselves to be able to accommodate this type of release, and we agree on its benefits, as it can provide an increase of quality assurance and reduce release time, with an overall benefit for the patients.