Determining Critical Quality Attributes Forms the Foundation for QbD Implementation

Since the principles of Quality by Design (QbD) for pharmaceutical manufacturing first emerged a few years ago, many companies have been struggling to understand the key concept of design space-what it is exactly, and how to establish it. The recent CMC Strategy Forum, however, started by examining a more fundamental issue: how to determine the critical quality attributes (CQAs) of a product-that is, the characteristics that define its quality.

“Critical quality attributes are the foundation of Quality by Design,” said Genentech’s Ron Taticek during his presentation at the conference, sponsored by CASSS (www.casss.org) and held in Bethesda, MD, on July 24 and 25. But if the three hours of presentations and discussions at the meeting were any indication, conducting risk assessments to determine the CQAs of a biotech product is not a simple task.

At the meeting, regulators and members of industry-about 140 people in all-discussed and debated the implications, risks, and benefits of various approaches to conducting such risk assessments, as well as various other issues involved in filing QbD data for biotech drugs, such as to how to establish a design space and then convey such information to regulators.

Taticek said he and his colleagues put a lot of thought into their risk-assessment approach, which has evolved over time. Initially, they tested the popular Failure Mode and Effects Analysis (FMEA) and Preliminary Hazards Analysis (PHA) methods. The team found both methods unacceptable, however, because they included occurrence and detectability factors, leading the team to underestimate risk levels. “Factoring in those two often meant taking things off the list [too] early, when not that much data was available,” he said.

Instead, the Genentech team chose to evaluate only the severity criteria, which they broke down into two factors: impact (on safety or efficacy) and certainty or uncertainty (that the attribute will have the expected impact). They then spent time revising the certainty scale. “That scale has changed numerous times since we started working on it,” he said, adding that the team is still refining its overall risk assessment approach.

Milton Axley of MedImmune and Andrew Weiskopf of Biogen Idec also gave talks at the session, and presented somewhat different approaches to determining CQAs. Axley’s risk assessment, for example, included both severity and likelihood factors. And whereas the Genentech and Biogen Idec groups divided attributes into only two types-critical and noncritical-the MedImmune approach added a third category, “key quality attributes.”

Another important difference in the methods was whether process capability-the ability to effectively control an attribute during manufacturing-was taken into account when deciding whether or not to classify an attribute as critical. Axley’s and Weiskopf’s methods took process capability into account, whereas the method presented by Taticek did not.

In a case study, Weiskopf showed the actual results that led his team at Biogen Idec to categorize attributes as critical or noncritical, and how process capability was factored in. Through structure-activity relationship studies, they evaluated what their process delivered with respect to attributes such as acidic isoforms and galactosylation, and then confirmed the impact of those attributes on bioactivity. For example, they examined Fc binding activity against a range of differentially galactosylated test articles, and established a range of comparable activity that was far wider than their process capability. Those findings substantially lowered their risk and enabled them to categorize galactosylation as noncritical.

Weiskopf said this approach helped his group prioritize their work. “Once you set aside the quality attributes which are clearly critical ... you can focus your efforts on gaining a deeper understanding of attributes whose impact on bioactivity is unknown or not fully quantified,” he said. These, he said, tend to be specific structural characteristics of the molecule, such as glycosylation, deamidation, or oxidation.

In the 90-minute discussion that followed the presentations, the group hashed out risks and factors that should be considered, and came to some broad conclusions.

Lifecycle Approach to CQAs Needed

One of the group’s main conclusions was that determining CQAs requires a lifecycle approach that evolves as knowledge about the product increases during clinical development and even commercial manufacturing.

“Early on, we will depend on prior knowledge from other products to come up with a list of potential CQAs,” said Axley. “We might even want to say that we consider them guilty until proven innocent. It’s an evolving and ongoing process.”

Taticek agreed, noting that even after a product is in commercial production, it may be necessary to update the CQA risk assessments. “You need a system that allows you to go back and reevaluate the original parameters you thought were important,” he said.

CQAs Do Not Necessarily Equal Specifications

Another important conclusion was that one should not necessarily expect to see a one-to-one relationship between CQAs and specifications. One reason for this is that several product variants identified as CQAs may be detected by a single test method and therefore built into a single specification.

Also, some attributes might be included in specifications even if not considered critical. This would occur if they were deemed to be a measure of process consistency, or conducted to meet a compendial or regulatory requirement, such as an identity test. Anthony Ridgway of Health Canada suggested coining a new term, “standard quality attributes,” for this category of attributes.

Conversely, some CQAs may not require a specification if they are well controlled or cleared within the process and can be “validated out.” Examples might include residual DNA, Protein A, or host cell proteins.

Another important factor discussed was how CQAs might combine with each other. “We cannot forget that the variability around multiple noncritical attributes might impact CQAs, or taken together, might have a direct impact on efficacy or safety,” noted Tony Mire-Sluis of Amgen, during his summary of the discussion.

Manufacturers Must Explain Criteria to Regulators

It also was agreed that it is the manufacturer’s responsibility to demonstrate to regulators what is critical or not critical for a given product and process and thus define the control strategy. Simply showing regulators the numbers assigned during a risk assessment is not very informative, said Susan Kirshner, a reviewer with the US Food and Drug Administration. “We like to see what thought process you used,” she said.

QbD Pilot and Industry Mock Case Studies Will Help Develop Concepts

How CQAs are used to shape a design space and support QbD filings were discussed during the other three workshops at the meeting. These questions also will be explored in greater depth in the FDA’s upcoming QbD pilot program for biotechnology drugs. The biotech pilot, announced on July 1, follows a small-molecule QbD pilot that was completed last year. See the FDA announcement at: http://www.fda.gov/ohrms/dockets/98fr/fda-2008-n-0355-n.pdf

In addition, mock QbD case studies for biotech products are being conducted in the US and Europe. By discussing a mock QbD filing for a theoretical biotech product, participants from different companies can share ideas freely.

Conformia, a software company that has a cooperative research and development agreement (CRADA) with the FDA, is leading the US exercise, which began on Aug. 5, with participation from seven companies: Amgen, Abbott, MedImmune (part of AstraZeneca), GlaxoSmithKline, Eli Lilly, Pfizer, and Genentech. Conformia led a similar mock case study last year for a fictitious small-molecule product, “ACE.” The results of the ACE case study are available on Conformia’s web site (www.conformia.com).

The European case study, which began earlier this year, is being conducted through the European Federation of Pharmaceutical Industries and Associations (EFPIA). It will describe the content of Section 3.2.S.2 of the common technical document (CTD) following the principles of the ICH Q8, Q9, and Q10 guidelines. EFPIA also conducted a small-molecule case study in 2005 of the P2 section of the CTD format (the “Examplain” study). A discussion and presentation of the Examplain study are available on the EFPIA web site at: http://www.efpia.org/Content/Default.asp?PageID=450.

Although many questions remain, the open discussions at this meeting helped industry and regulators increase their understanding of how to apply the principles of QbD. John Dougherty of Eli Lilly said he was very pleased to see how much progress had been made since the first CMC Strategy Forum on QbD a year ago. “I’m amazed at the clarity of discussions this year compared to last July,” he said. “In a very short time, we’ve come a very long way.”