OR WAIT 15 SECS
Second in a three-part series that discusses the complexities of QbD implementation in biotech development.
The International Conference on Harmonization (ICH) Q8(R2), Q9, and Q10 guidelines provide the foundation for implementing Quality by Design (QbD). Applying those concepts to the manufacture of biotech products, however, involves some nuances and complexities. Therefore, this paper offers guidance and interpretation for implementing QbD for biopharmaceuticals, from early-phase development steps such as identifying critical quality attributes and setting specifications, followed by the development of the design space and establishing the process control strategy; to later stages, including incorporating QbD into a regulatory filing and facilitating efficient commercial processes and manufacturing change flexibility post licensure.
This paper focuses on the factors to consider when applying the Quality by Design (QbD) concepts outlined in ICH Q8(R2), Q9, and Q10 to biotechnology products. Although biologic and biotechnology products often present a higher level of complexity than small molecules in terms of manufacturing process or product structure, the concepts of QbD are the same as those for small molecules. This paper describes the nuances and complexities involved in implementing QbD in the manufacture of biotech products and offers guidance and interpretation for doing so. The scope of this paper is limited to well-characterized protein products, in which the natural molecular heterogeneity, impurity profile, and potency can be defined with a high degree of confidence.
(Eli Lilly and Company)
Part 1 of this three-part article, which appeared in the November issue, covered molecular design, the use of laboratory and clinical studies to identify critical quality attributes, setting specifications, and developing the design space. Here, in Part 2, we address the use of design of experiments (DOE) to define the design space, unique considerations for process development for biopharmaceuticals, the establishment of a control strategy, and the placement of QbD information in a regulatory application.
Process parameters that have been identified, through risk analysis, as potentially having a significant impact on subsequent process steps, may be studied using multifactor DOE. DOE is more efficient and effective than traditional "one-factor-at-a-time" (OFAT) experiments. DOE is more efficient because it (1) requires fewer numbers of experimental runs, and (2) covers a broader "knowledge space" than OFAT experimentation. As a result, it is more effective in (1) investigating potential interactions among process factors, (2) avoiding artifacts such as experimental clustering and run order through randomization, and (3) making use of "hidden replication," and thus in having better sensitivity for detecting important effects.
The mathematical model that is derived from DOE can be used together with the acceptable boundaries of critical quality attributes (CQAs) to define a design space for a given process step. This is illustrated in Figure 1. Two response factors (X1 and X2) are studied across the knowledge space that has been defined by the multifactor DOE. These factors yield a response surface for a CQA (Panel 1). The response surface intersects the lower (Panel 2) and upper (Panel 3) specification limits (USL and LSL) for a subsequent process step, to yield its design space (Panel 4). That space represents the normal operating ranges for the factors, falling well within the design space (Panel 5). Operating within this normal operating range (NOR) will yield quality attribute measurements that fall within the upper and lower control limits (UCL and LCL in Panel 6). Because the LCL and UCL fall well within the LSL and USL, the process step is predicted to be highly capable of delivering product that meets the requirements of subsequent steps in the process. Excursions outside the NOR are expected to deliver product with quality attributes that are acceptable for further processing, as long as the operating parameters are held to limits defined by the design space.
Figure 1 illustrates the design space for a single quality attribute. In biotech processes, however, it is typical for a single unit operation to affect several quality attributes. For such cases, the design space for the unit operation is obtained from overlays of the design spaces derived from analyses of multiple attributes, or from a multivariate analysis of the system. In addition, the design space is derived from a mathematical model that is subject to uncertainty and to the limits of the model. With continual verification and updating of process information through information management, the uncertainty is reduced because more data are gathered and included during the lifecycle of a product.
Process development and characterization studies are primarily performed at laboratory scale to make them cheaper and faster. Therefore, developing a representative scaled-down model is crucial to meaningful process characterization and successful process-fit analysis. For example, scale-down of a fermentation step may involve keeping the scale-independent parameters (such as process temperature, pH, inoculation percentages for each step, and times of feed media additions) at the same control set point as the large-scale process, assuming similar vessel geometries, and changing the mixing and air flow rate to mimic the conditions at large scale. Qualifying such scale-down models before their use in process characterization studies is key to establishing a representative design space.
Before defining a cell culture process, the manufacturer may perform a clone selection (in the case of mammalian or microbial expression systems). A suitable clone must be selected from a number of available clones for establishment of the master cell bank. The clone selection should be based on the quality target product profile (QTPP) and a risk assessment. Parameters determining the selection will be a combination of product quality attributes (e.g., impurity profile) and process performance (e.g., specific productivity per cell, consistent yield), usually assessed at small scale. Prior knowledge of process parameter ranges, existing cell culture platforms, and literature data will play a significant role in clone selection. This could be combined with a mechanistic approach. Alternatively, DOE could be applied for a multivariate approach if a larger number of parameters are investigated.
A fundamental part of cell culture process development is the development of optimal cell culture media for the different stages of cell culture (e.g., inoculum preparation, main stage cell culture). This can be achieved using small-scale models. The impact of cell culture media composition can be identified by using tools such as DOE. This also provides information that supports the definition of a design space that includes critical attributes of raw materials. Optimal cell culture media development also may include the testing of critical raw materials (supplements) from different suppliers to develop appropriate raw material specifications to allow for alternate suppliers.
For cell culture unit operations, a risk assessment should be performed to select the process parameters and the drug substance or drug product quality attributes to be investigated in a systematic approach (e.g., DOE). The risk assessment will take into account the type of host cell and expression system, prior knowledge, platform technologies, and knowledge from the clone selection and literature data, and should include raw materials and their impact on CQAs. Process parameters selected are those that may have an effect on quality attributes that are linked to the identity, strength, quality, purity, or potency of the product and may ultimately relate to the safety and efficacy of the product. In addition, those process parameters are commonly investigated that affect various aspects of process performance, such as product yield. There may be cases when CQAs cannot be measured in the cell culture broth because of interference from other components present in the sample matrix or other limitations associated with the analytical assay. To evaluate the influence of cell culture process parameters on quality attributes for such cases, it may be necessary to process the output from some of these small-scale experiments over the subsequent process steps until adequate product quality determinations can be made.
The process steps are tightly linked, with the output of the previous step serving as the input to the next step. The focus in cell culture is on consistency, productivity, product integrity, and freedom from contamination. The focus of purification is on purity and freedom from process and related materials.
For drug substance purification and drug product formulation steps, the same principles, as outlined for cell culture, apply for the establishment of a design space. Whereas in cell culture, the individual steps of the process differ mainly in terms of scale, container type, and culture media, each purification unit operation may comprise a distinct technology. This situation provides the opportunity to establish a design space for each unit operation, as well as to combine several different unit operations in a certain order to ensure the desired quality of the product.
The ICH Q10 definition for control strategy indicates that controls for a product consist not only of process controls and final specifications for drug substance and drug product but also controls associated with the raw materials, excipients, container and closure, manufacturing equipment, and facility. It is a state of control in which all of the "planned controls" work together to ensure that the product delivered to the patient meets the patient's needs. Design space boundaries, as described above, are an integral part of a comprehensive control strategy. The control strategy for a product is expected to evolve through the product lifecycle. As product knowledge evolves, there is the likelihood of less reliance on end-product testing.
The purpose of a control strategy for a product is to ensure that sufficient controls are in place to maintain the risks associated with the product at a tolerable level. Thus, the concepts of risk management and control strategy are intimately connected. Product risk is synonymous with potential risk to the patient. The risk from a product is measured in terms of the amount of harm (or losses because of the inability to supply the market) it can potentially cause. This harm is created by hazards that are associated with the product and its manufacture and supply. The control strategy builds in layers of protection that reduce the risk of the hazards creating actual harm. Given that each layer of the control strategy will not be perfect, it will always be possible for hazards to find a way through the controls to create harm. This concept is illustrated by the "Swiss Cheese" model of risk shown in Figure 2.1 Each control strategy layer can have holes, and if the holes line up, a hazard can become a harm (or loss). The following are four considerations that must be addressed when evaluating the ability of a control strategy to manage risks:
1. Are there enough layers of protection in the control strategy?
2. Are there some layers that are not required because they do not protect against any harm?
3. Does each layer of protection function adequately?
4. Does the control strategy itself introduce any new hazards?
This integration of risk management and control strategy can be done in conjunction with the principles described in ICH Q9. Coupled with the principles of QbD defined by ICH Q8 and the appropriate Quality Systems (ICH Q10), the effective use of risk management will dictate a sufficient level of controls to protect the product without introducing new failure modes (for example, from inappropriate product sampling) or without incurring excessive cost that is passed along to patients. A well-designed control strategy that results from appropriate leveraging of Q8/Q9/Q10 principles, then, leads to reliable product quality and patient safety profiles. Although firms may not choose to leverage these principles in exactly the same way, all effective product control strategies use at least some of the elements of risk management.
For biotechnology products, a holistic approach to control strategy is of particular importance because the complexity of biopharmaceutical products makes it unlikely that product quality can be adequately controlled simply by ensuring that end-product specifications are met. Although it has often been said that the quality of a recombinant product is dictated by the process rather than driven by specification testing, with QbD the quality is designed into both the product and process, and a combination of in-process controls and end product specification testing is used to verify that quality. The integral link between process and product defines unique challenges for the development of robust control strategies for biotechnology products. Some of these challenges for biotechnology products are:
A holistic approach to the control strategy helps minimize the probability of a negative impact on product safety and efficacy. In addition, establishing a comprehensive control strategy can further reduce risk by taking the following elements into consideration:
Consider a monoclonal antibody (MAb) that was derived from a Chinese hamster ovary (CHO) cell culture and directed against rheumatoid arthritis. The QTPP for this product indicates that a typical patient needs the product to be delivered in a pen that may be held at room temperature for at least one week. A stable single-use solution drug product using a pre-filled syringe for once-weekly dosing meets the QTPP. The patient population typically has reduced manual dexterity and the pre-filled syringe means there is less need for patients to perform manual product manipulations before dosing (i.e., the product has been designed to meet patient needs). The amino acid sequence for the selected MAb with the greatest efficacy has a methionine in the complimentarity determining region (CDR). Oxidation of this residue in the drug product results in loss of potency, which affects product efficacy. Purity with respect to oxidation, then, is a CQA for this drug product. The formulation design and manufacturing control strategy must be developed for this CQA to minimize the oxidation of the methionine residue during manufacture and shelf storage. Appropriate preformulation and formulation DOE studies must be conducted to define the variables affecting the oxidation of methionine, such as pH or dissolved oxygen (DO), or peroxide or trace metals from raw materials or container closure systems. These studies lead to the identification of the design space where product stability is acceptable. Additional controls define the appropriate control strategy during manufacture as well as for incoming excipients and packaging components. For example, the presence of peroxides in excipients (e.g., surfactants) or potential leachates from packaging components can increase the susceptibility to oxidation and require careful monitoring (e.g., vendor qualification) and control. Controls in manufacturing such as monitoring levels of DO using appropriate in-line process analytical technology (PAT) measurements and temperature controls may be instituted to control the oxidation and stability CQA to obviate an end testing specification. Similarly, processes should be designed such that every batch delivers a safe, efficacious product, consistent with the product design intent outlined in the QTPP.
Development of biotechnology products based on the QbD principles presented in this paper will require modifications to the current application format and content. It is expected that the standard common technical document (CTD) format, for the most part, will be retained in a "QbD" application or supplement but that the content of specific sections in Module 3 will reflect "knowledge rich" information required to support a QbD biological product. The extent of these modifications will depend on how much QbD information individual sponsors may wish to present in their applications. For the purposes of this discussion, we have assumed the sponsor will be maximizing the amount of QbD information to cover manufacture and testing of both the drug substance and drug product. Sponsors may decide to limit QbD to the drug substance, the drug product, or even individual unit operations of either the drug substance or drug product.
There are a number of principles one must consider when preparing and optimizing quality information in the application. In addition to providing a comprehensive "reviewer friendly" data package that meets regulatory requirements, a strong focus on the data required to establish the relationships between the CQAs and the resulting control strategy for drug substance and drug product is essential. This focus should also include the design of key development studies and the risk assessments that were considered in the design of these studies, as well as a summary of the key findings and conclusions. Every effort should be made to avoid the tendency to include every piece of data to support one's conclusions. "Knowledge rich" is not a data dump but a more focused, strategic presentation of the data synthesized into information to allow a comprehensive review and understanding of the product and manufacturing process.
The following section presents a recommendation for incorporating QbD information into a CTD format. A sponsor might choose to apply some or all elements of the recommendation. The regulatory landscape for QbD applications will likely continue to evolve in the future.
Module 1: Administrative Information. The proposed post-approval management plan may be included in this section.
Table 1. Suggested placement of QbD information in the common technical document (CTD) for the drug substance
Module 2: Quality Overall Summary. The role of the Module 2 Quality Overall Summary (QOS) in a "QbD" application is not expected to differ significantly from the QOS in a more traditional application, however, there is the opportunity to consolidate or cross-reference data that comprehensively describes the design space and control strategies.
Module 3: Drug Substance and Drug Product. Tables 1 and 2 provide possible locations for the QbD information about the drug substance and drug product. Sections that do not contain QbD information and contain only information that would be included in a traditional application are identified with the term "traditional content." In this proposal, section 3.2.S.2.6 (Manufacturing Process Development) is the key QbD section for drug substance and Section 3.2.P.2 (Pharmaceutical Development) is the key QbD section for drug product.
Table 2a. Suggested placement of Quality by Design information in the common technical document (CTD) for the drug product (DP)
Part 3 of this article will appear in the January 2010 issue. It will discuss continuous verification and post-approval changes, including topics such as verification at large scale, refinement of the design space, process changes and comparability, comparability protocols and expanded change protocols, marketed products, and a CMC post-approval management plan. It will also include the overall conclusions.
Table 2b. Suggested placement of Quality by Design information in the common technical document (CTD) for the drug product (DP), continued.
Taruna Arora is a principal scientist, protein science, Roger Greene is the director of regulatory affairs, Jennifer Mercer is the director of regulatory affairs, CMC, and Paul Tsang is the executive director of quality, all at Amgen Inc.; Meg Casais is the director of global regulatory affairs, CMC, and Stuart Feldman is the director/CMC, global regulatory affairs, both at Schering Plough; Jutta Look is the senior manager, CMC Regulatory Affairs at Novartis Pharma AG; Tony Lubiniecki is the vice president of biopharmaceutical development & marketed product support at Centocor R&D, Inc.; Joseph Mezzatesta is the assistant director of regulatory CMC, corporate regulatory affairs, at Sanofi-Aventis; Stefanie Pluschkell is an associate research fellow in global CMC, biologics and devices, at Pfizer Inc.; Mark Rosolowsky is the executive director of global regulatory sciences–CMC at Bristol Myers Squibb; Anurag Rathore is a biotech CMC consultant and faculty member at the Indian Institute of Technology, Delhi, India; Mark Schenerman is the vice president of analytical biochemistry at MedImmune; Tim Schofield is the director of US regulatory affairs at GlaxoSmithKline; Samantha Sheridan is the director of regulatory affairs at Shire Pharmaceuticals; Paul Smock is a director and quality product leader in Wyeth Biotech, Wyeth Pharmaceuticals; Sally Anliker is the director of regulatory affairs CMC, Lois Atkins is a principal consultant, CMC regulatory affairs, Bernerd McGarvey is an engineering advisor in the process engineering center, Bruce Meiklejohn is a principal fellow, regulatory COE-biotech, Jim Precup is a research scientist in manufacturing, science and technology, and John Towns is the senior director of global CMC regulatory affairs, all at Eli Lilly and Company. Towns is also the chair of the working group, 317.276.4079, Towns_John_K@lilly.com
1. Reason J. Human error: models and management. BMJ. 2000 Mar 18;320(7237):768–70.
2. Glodek M, Liebowitz S, McCarthy R, et al. Process robustness—A PQRI white paper. Pharm Eng. 2006 Nov/Dec;26(6):1–11.
3. Parenteral Drug Association. PDA Technical Report 42. Process validation of protein manufacturing. 2005 Sep;59(S-4):1–28. Bethesda, MD: PDA.