QbD Update: More Science-Based Submissions but Also More Questions and Confusion

Industry and regulators agree that they are still in the learning mode when it comes to implementing the quality-by-design paradigm for developing and producing pharmaceuticals.

The pharmaceutical industry and regulators are continuing to grapple with a variety of challenges and are still in the learning mode in implementing the quality-by-design approach to manufacturing.

Some of the ongoing challenges identified at a Jan. 28-29 Joint Regulators/Industry QbD Workshop in London sponsored by the European Medicines Agency and the Parenteral Drug Association include the inconsistent use of terminology, and heavy workload and data requirements.

Conferring regulatory relief was one incentive that regulators offered to manufacturers in exchange for QbD submissions, but lack of clarity about the specific relief manufacturers would get has frustrated manufacturers.

At the meeting, six case studies were presented showing how QbD is working in practice, where industry representatives and assessors described some of the lessons learned.

Mixed picture

There is a mixed picture on whether uptake on QbD is growing. While the number of QbD applications in the European Union is down, it is growing in Japan.

The status of QbD submissions is unclear in the U.S. The FDA does not track QbD submissions, saying it is too difficult to do so. FDA officials say that QbD should be viewed as a continuum and that all applications have some aspects of QbD (Also see "Slogging Toward Quality by Design" - Pink Sheet, 20 Dec, 2012.).

Yoshihiro Matsuda of Japan’s Pharmaceutical and Medical Devices Agency (PMDA) told the meeting that after averaging approximately three QbD applications annually for a few years, the number jumped to 11 submissions per year in 2011-2012 and six submissions by mid-2013.

Jean-Louis Robert, head of the unit of the National Health Laboratory Medicines Control Service in Luxemburg, said the number of QbD applications submitted in the EU dropped in 2013 from 2012.

Stage 2 of the learning curve

Georges France, head of external affairs for Novartis, said that the pharmaceutical industry has arrived at a critical juncture in what he described as a three-stage process of QbD implementation.

The first stage was achieving process understanding, efficiency and yield improvement, and reducing out-of-specification results and recalls through adoption of QbD.

The second stage involves additional training of inspectors and industry on QbD, the shared pilot program.

The third stage involves regulatory trust and global harmonization of QbD.

France said that “we have not yet achieved Step 3. This is where trust and predictability come in. This is where regulatory flexibility comes, and global regulatory alignment and continuous improvement and post-approval changes.”

“The second step is where we are today. What we are looking for is the full benefit.”

More science yet more complaints

In the opening session, Robert seemed to agree with France that the industry was still firmly planted in Step 2 and is still in a learning mode and that there are still many questions associated with QbD.

He said that while QbD has resulted in more science-based applications, with more scientific understanding and the use of design of experiments and more robust manufacturing processes, there have also been more questions from both regulators and industry alike.

“There is more science but more questions. The level of data and information requested or submitted in an application is not really clarified.”

“There are also complaints that the term QbD is unclear and is sometimes misused. Also there is inconsistent use of terminology.”

Robert also hears complaints from industry questioning why they need to do QbD if “there is no benefit at the end.”

The International Conference on Harmonization has formed an informal Quality Discussion Group to identify and refine guidance on a range of QbD topics.

One of the things that the group is looking at is to define where QbD information such as design space or chemometrics or statistical modeling should be located in the Common Technical Document.

Graham Cook of Pfizer said that “Pfizer uses QbD concepts as a way of life now for developing new products. We get a significant benefit in terms of improved robustness as opposed to products developed through more traditional approaches.”

Yet Cook also noted that “what we have seen recently is that there is significant effort in putting together QbD submission. It involves a lot more work. The impression is that this can be challenging for assessors as well.”

Industry has come a long way

Christine Moore, the acting director of FDA’s Office of New Drug Quality Assessment, said that the pharmaceutical industry has made much progress in implementing QbD over the past 10 years.

She said that overall “I think it shows that we have made a lot of progress with QbD implementation. … I think the question has changed. If we look at where we were eight to 10 years ago, people were asking the question should I do QbD? And now I think the question more is how I should do QbD or should I register my QbD as part of the application? But it has evolved and the value for doing the QbD at least in terms of your development I think the case has been established for itself. And that is great because that puts us on common ground. These are good and valuable approaches to move forward and now it is just sorting through the details on how to optimize that.”

Moore concurred with her colleagues, however, that more work needs to be done in resolving what is meant by certain terms and definitions.

She said that certain themes cropped up repeatedly in the case studies presented at the meeting. These included the following:

• Classifying criticality;
• Level of detail in process descriptions;
• Design space verification;
• Level of detail in risk assessments;
• How to change non-critical process parameters; and
• How to summarize control strategy.

She said that the first three items have been addressed by EMA and FDA regulators in two recent guidance documents issued under the joint QbD review program (Also see "FDA and EU Extend QbD Pilot Program and Continue to Clarify Expectations for QbD" - Pink Sheet, 30 May, 2014.).

Moore said that the fourth area, the level of detail on risk assessments, will also be taken up in a forthcoming guidance about to be issued under the joint QbD review program.

These six themes, she said, are all related to post-approval changes. It is now up to regulators to figure out what to do with some of this QbD information in the filings.

Moore said that “in the last 10 years or more we challenged the industry to put quality- and risk-based approaches and science into your pharmaceutical development and manufacturing and you have responded and you have done that quite a bit. Now it comes back to the regulators to say, ‘alright we have all this good information, how can we take advantage of it not only in the initial filing but throughout the lifecycle of the product?’ This has been a great dialogue to help move that next step forward.”

Definitional challenges cited

Moore also noted some other definitional challenges that are blocking fuller QbD implementation.

Moore said that “we have a lot of definitions, and I was sitting back in the room and I realized that people are not using the same definitions.”

Moore described some examples:

• Non-critical process parameters. Moore said that “we have had enough difficulty interpreting the ICH definition of critical. This is not an ICH term and is not defined anywhere. Yet different people talk about different things. What do you mean by a non-CPP? Does this mean this does not have the potential to affect critical quality attributes anywhere for any range or within the ranges studied? This might sound petty but as we are looking to figure out if we are to do something with non- critical process parameters and provide regulatory flexibility we have to be clear with what we’re talking about.”
• Proven acceptable ranges. Moore said that “there is a definition in ICH but I would bet that most of you who use PARs are not using that definition. The definition in ICH said that PARs are based upon univariate experiments. I have heard not only at this meeting but at others three or four different definitions of PARs. I have heard if you do a design of experiment and if there are no interactions, that’s a PAR. That’s an interesting definition. … I have heard other definitions today. So if you’re going to us PARs be careful with how you’re using them.”
• Model maintenance. Moore said that model maintenance uses both the terms out of trend and out of specification. “I heard a lot of confusion yesterday about out of specification and out of trend outliers especially for chemometric type models. I am not going to go into details of that here but I want to point out that this is an important point to understand. I know for many people models are really scary. I understand that, and that you have technical people who do all of these aspects. But I really believe that these sorts of aspects of how you implement models this all affects your quality systems and the decisions you’re making on a daily basis. If you don’t personally understand what I’m saying by out of spec or out of trend outliers you should probably study up on it.”

Study shows risk assessment conflict

Some of the disparate interpretations and definitional challenges with respect to QbD were evident in some of the case studies presented at the meeting.

Frank Montgomery of AstraZeneca said that the company submitted two QbD applications for small molecule immediate-release tablets in 2010 and 2011. Montgomery explained that the company wanted to use QbD for their submission “to understand our product and process robustness.”

With “Product A” they wanted to reduce their reliance on end-product testing. The company submitted a complex design space for both the API and the product. This was submitted in 2011.

“Product B” involved a more traditional overall control strategy. There were discrete design space proposals for both the drug product and the drug substance. This was submitted in 2010.

The company wanted to have wider process parameter ranges were supported by design of experiments.

Montgomery noted that “with one of our products we had a lot of questions and a lot of data, as Jean Louis mentioned. But maybe this was justified? With one of our products there were a lot of questions and a lot of data was required to support the proposal. It was a huge challenge not just for the company but for the assessors as well that had to go through that data as well. I’m not sure that’s where we want to be. “

EMA did not like traffic lights

Montgomery noted that regulators did not like the “traffic light” approach that the company used to present risk information for the products. The company presented the information in a chart format only with no additional descriptions.

Low risk was assigned a green color, medium risk was yellow and high risk was red. Montgomery said that the regulators did not like the traffic light color format because they felt it was overly simplistic.

Montgomery said that regulators wanted to see the underlying data to support these risk assessments. “There is a huge number of raw data that we generate that is difficult to summarize.”

Yet Montgomery pointed out the challenges of supplying regulators with the voluminous amounts of raw data that were used to support the risk assessment as well as the fear that the data could be misinterpreted.

Regulatory take on risk assessments

For her part, Olvia Lake, an EU quality assessor who reviewed the AstraZeneca application, agreed that regulators want to see more detail in risk assessments.

She said that often “only a summary table is presented without explaining how the risks have been classified. This is not sufficient. The risk assessment tool should be stated and scoring and thresholds used to classify the risks should be explained.”

Lake also advised that manufacturers should ensure that all known risk factors, such as degradation, have been included in the risk assessment. Furthermore, risk ranking outcomes not in line with existing scientific knowledge should be justified.

Lake also offered the following pointers on how to conduct effective risk assessments:

• The link between risk assessment and drug substance/drug product specification should be clear and the absence of potential critical quality attributes in the specification should be justified.
• Manufacturers should ensure that the risks are managed by the design space or the proposed control strategy.

Lake said that a good example for a risk assessment table is presented in the training material of the ICH Q-IWG on the implementation of the ICH Q8, Q9 and Q10 trio of guidelines.

Lake also noted that “you cannot work with the risk assessment until the Quality Target Product Profile (QTPP) is presented. We really want to see the QTPP for the enhanced approach so that we know what you’re doing.”

Lake also said that “we pay a lot of attention to risk assessments.”