FDA Reports Growing Use Of Process Modeling Tools To Support Continuous Manufacturing

FDA reports “growing traction” of continuous manufacturing and a corresponding uptick in the use of models to support this mode of manufacturing. Officials explain how they assess these models during inspections.

FDA officials report growing interest in the pharmaceutical industry in the use of process modeling tools to support continuous manufacturing. Sponsors are also indicating in meetings with FDA that these models play a “significant role” in their control strategy for continuous manufacturing.

At a series of recent industry meetings, FDA officials gave an update on the status of continuous manufacturing, the industry’s use of these models, and the criteria the agency uses to assess these models.

The agency has an interest in promoting continuous manufacturing and other new technologies that can improve product quality, minimize product defects and reduce drug shortages – and has been championing them for over a decade.  (Also see "Continuous Manufacturing Poised to Disrupt Pharma Sector" - Pink Sheet, 29 Jul, 2014.)

The need for modeling tools is critical in continuous manufacturing because the process cannot be stopped, unlike batch processes, and these tools can predict whether processes can remain in a state of control. In batch manufacturing, processes are stopped after each unit operation and products can be evaluated after each step.

Growing Traction Of Continuous Manufacturing

FDA’s Sharmista Chatterjee told the American Association of Pharmaceutical Scientists meeting on Nov. 6 in Washington that the agency is “seeing a lot of traction in continuous manufacturing.” Chatterjee is the director of the division of process assessment II in the Office of Process and Facilities in the Center for Drug Evaluation and Research.

She said that there is also growing interest in developing process modeling tools to support this mode of manufacturing.

FDA’s Sau Lee supported his colleague’s claims on the growing interest in continuous manufacturing. He said that there has been a sizeable number of requests to join the FDA’s Emerging Technology Team (ETT), signaling an increase in continuous manufacturing applications awaiting FDA approval.

Lee presented on the work being done by the FDA’s ETT to promote continuous manufacturing on Nov. 6 at the International Society for Pharmaceutical Engineering annual meeting in Philadelphia on. Lee is the director of ETT.

Lee said that there have been over 30 requests accepted into the ETT since the launch of the program in late 2014. In addition, there have been over 60 ETT-industry interactions including both teleconferences and face-to-face meetings, with about half of these discussions focused on continuous manufacturing. ETT was formed in 2013 to promote the use of new technologies in pharmaceutical manufacturing.

The number of requests is significant since the agency only accepts requests if they are from firms that are planning to submit investigational new drug applications, original or supplemental new drug applications, abbreviated new drug applications, biological license applications or applications associated drug master files that involve emerging technologies.

Lee said that sponsors are indicating in these meetings that these models play a “significant role” in the control strategies they are developing for continuous manufacturing.

FDA’s first approval of a continuously manufactured drug came in July 2015 for Vertex Pharmaceuticals Inc.’s Orkambi; the next was in April 2016 for Janssen Pharmaceutical Cos.' supplemental new drug application switching Prezista from a batch to a continuous process. Since then, two more applications have been approved: Eli Lilly & Co.'s  Verzenio and Vertex’s Symdeko, both in February 2018.

Three Models Used For Continuous Manufacturing

Chatterjee said the industry is using three types of models in continuous manufacturing: the mechanistic model, the empirical model, and a hybrid model which combines the two. She said that “in the continuous manufacturing space we have seen all three categories used.”

Mechanistic models employ the use of population balance model which can be used to measure the powder flow in the continuous blender. Empirical models are associated with the near-infrared (NIR) method and multivariate data analysis models for process monitoring. The hybrid model combines both the mechanistic and chemometric models.

Chatterjee said that a regulatory framework was needed to help address industry’s questions on “what type of information should I provide about the model?” FDA also needed a tool to aid its regulatory decision making on evaluating whether these models adequately demonstrate that processes are in a state of control.

FDA uses a guideline developed by the International Council for Harmonization to help answer these questions, said Chatterjee. The “ICH Endorsed Guide for ICH Q8, Q9, and Q10 Implementation” was issued on Dec. 6, 2011. It categorizes models into three levels based on their risk, a low impact model and a high impact models and a medium impact model:

• A high impact model: predictions from the model are used as the sole indicator of quality of the product. Chatterjee said that an example of a high impact model is when the NIR sensor and its chemometric model is used for real time release testing.

• A medium impact model is used for assuring quality of the product but is not the sole indicator of quality. This can include the use of NIR to flag non-conforming product. Chatterjee said that an example of a medium impact model is when a model is used in commercial manufacturing to predict the active concentration and the diversion of non-conforming product.

• A low-impact model is typically used to support process development efforts. She said that an example of a low impact model would use residence time distribution to verify feeder limits, characterize powder mixing, or predict dispersion through the blender.

What FDA Looks For In Inspections

Chatterjee also described how the agency uses these models in preapproval inspections.

For the high impact model, “we want to get an understanding of your general idea for model maintenance.” FDA also wants to know the “trigger” for updating the model, the criteria for recalibration, and the level of validation of the model.

Yet Chatterjee cautioned against providing too much data during inspections.

She recounted an experience during a preapproval inspection of one firm that was using a high-impact model. Before the inspection, the company provided the FDA with a decision tree on its approach for maintaining the model. However, when FDA went to perform the inspection, the firm provided the agency with reams of documents on its standard operating procedures for model maintenance.

She said that “we don’t need to see all these details.” Rather the agency would like to see a summary of the model maintenance plan but not all the details.

Chatterjee said that it is also important during inspections to have the supporting data on hand to support the use of the model if requested.

One firm during an inspection did not have the information to support the proposed loss in weight feeder limits. FDA officials went back to their lab and ran a simulation to ensure that the feeder limits were correct. The lab found that the limits were acceptable so the agency did not have to submit another information request to the sponsor.

Chatterjee Says Agency Has Modeling Expertise

At the end of her presentation, Chatterjee was asked by a member of the audience whether FDA has the internal expertise to review the modeling supporting continuous manufacturing applications.

Chatterjee responded that “we do have modeling expertise, we have four full-time people who are modelers and we have access to modeling software and we have given out grants to academics and some of these models have been adopted.”

She was also asked if FDA is interacting with other regulators on aligning its expectations of continuous manufacturing. Chatterjee responded that FDA has met with regulators from Japan’s Pharmaceuticals and Medical Devices Agency and the EU’s European Medicines Agency. Chatterjee reported that “I am happy to say that in discussions with these main authorities I have to say we are quite aligned.”

Chatterjee added that an ICH Q13 working group was meeting in Charlotte NC in mid-November to being work on developing a plan to harmonize continuous manufacturing. ICH announced on June 22 at a June 2-7 meeting in Kobe, Japan, to create a Q13 guidance document on continuous manufacturing,  (Also see "The Quality Lowdown: ICH Q13 and Q14, Gene Therapy CMC, Drug Shortage Gaps, And An Excipient Misstep" - Pink Sheet, 25 Jun, 2018.)

Need For Alignment With Other Regulators

FDA’s Sau Lee picked up on the theme of global alignment as well as areas of future collaboration with industry in his remarks at the ISPE meeting.

He said that there is a need for global alignment on how to maintain and update these models. This topic of model maintenance could foreseeably be tackled by ICH Q13 as well as the upcoming guideline ICH Q12 on postapproval changes.

“I feel there is a great opportunity to do something to achieve alignment in this area. These models and updates may not be consistent among regulators, every time the industry makes an update to the model they may need to submit a supplement. I think what the industry wants is a practical way to manage these updates.”

Another potential area for collaboration with industry the topic of data management and storage.

He said that “because of continuous manufacturing we can obtain an unprecedented amount of process data to drive different process understanding and process improvement. However, this does pose some challenges in terms of data management and data storage. Do we need to store all this data? Can we just share some of the key data to enable us to perform an investigation in case of a recall or an adverse events? We are open to talking to industry about this area.”

From the editors of the Gold Sheet.