Filip Vanhoutte is an associate director of drug-product development at Johnson & Johnson Pharmaceutical Research and Development, Beerse, Belgium, tel. +32 0 14 60 39 58, fax +32 0 14 60 5333

Filip Vanhoutte

At the end of the 1990s, many pharmaceutical companies embarked on Six Sigma programs that brought chemometric tools such as design of experiments (DOE) and risk-analysis tools such as failure mode and effects analysis (FMEA) to developers. The Current Good Manufacturing Practices (GMPs) for the 21st Century initiative in 2002, the process analytical technology (PAT) guidance, and the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use's ICH Q8 

 have accelerated the acceptance of these engineering tools (1–4). 

However, developers were initially confused about how to implement ICH Q8. Scientists wondered how to put the guideline into practice in the diverse product-development teams spread around the globe, whether one could only realize quality by design (QbD) by applying PAT, how a product team could track how a project moves toward realization of QbD, and how to ensure that knowledge increases after a product is launched. 

As the concepts of QbD began to mature, the US Food and Drug Administration's Office of New Drug Chemistry (ONDC) announced a formal pilot program to provide additional information to help ONDC implement a new quality-assessment system to facilitate the submission and review of QbD in applications (5). The aforementioned tools found widespread use during the past decade. On the other hand, the integrated use of these tools in the research-and-development environment and their operational use in manufacturing settings have only been explored during the past several years. Similarly, the application of these tools in regulatory filings has evolved and continues to evolve in the regulatory, development, and manufacturing environments, mostly aided by exercising these practices as in the pilot program. 

The authors participated in the QbD pilot program for a postapproval change: an additional active pharmaceutical ingredient (API) manufacturing site and additional strengths of a product that had been granted accelerated approval one year before. The authors' contribution to the pilot program was not in the application of complex DOEs or PAT. Rather, it showed new ways of tracking the product knowledge, process, and material understanding, and how risk analysis is used to designate which process parameters or material attributes are critical and which are not.  

If QbD is presented in too complex a way, one will struggle to implement it. It is therefore important to keep the message simple, find the right emphasis, and offer development teams the key tools to guide them along the path to QbD. Key tools include a target product profile, a multigeneration plan, a concept-selection matrix, prior knowledge of formulation and process platforms and materials, a project risk analysis, criticality management, a development plan, DOE, and PAT. Other systems such as technical-design review meetings, development-deliverables checklists, and stage-gate decision meetings ensure and control the progress of the project. 

In addition, the development team must have scientific and quality-engineering or chemometrics competencies. When combined with scientific principles and relevant measurement responses, DOE is the most important tool for understanding causality in an efficient and multivariate way as well as the related factors and interactions. In contrast, the analysis of historical data (even from as many as 100 batches) that were not set up in a designed way, will never provide certainty about which factors are responsible for the observed variation. Most of the time, factors are confounded and show chance correlations. Initially, even PAT should once be combined with a DOE if one must understand causality. 

Another important tool is criticality management, which the authors define as a systematic approach to define the critical quality attributes of the drug product; the critical variability in formulation, packaging, materials, and the manufacturing process; and how critical variability can be optimally controlled. Criticality management combines pharmaceutical product, process, and material knowledge (required by ICH Q8) and risk management (required by ICH Q9) in one approach, which is reflected in a single document. Just as in the risk-management process, criticality management assesses, controls, communicates, and reviews criticality. During process-technology transfer, the criticality-management document is transferred to manufacturing operations and updated with new knowledge that is acquired during technology transfer. This update will also happen after approval throughout the entire product life cycle. 

Criticality management: an integrated approach

The literature describes the criticality management of the API-synthesis process (6). This article describes the criticality management of the drug product and its manufacturing process using a direct-compression tablet as an example (see Figure 1). 

Figure 1 (ALL FIGURES ARE COURTESY OF THE AUTHORS.) 

The drug-product criticality-management process translates  the target product profile into a list of potentially critical quality attributes of the drug product. The formulation is designed and selected based upon these quality criteria and further optimized by studying and managing the criticality of the composition and material properties, the packaging design, shipping, stability, and the in-use procedure. 

Translating the target product profile into critical quality attributes (CQAs) is the initial step. It is important that the right attributes be defined with proper measurement methods. These quality attributes are initially chosen based upon the requirements of the target product profile, prior knowledge of this type of product, the manufacturing process, and the sensitivities of the API. The attributes are further refined during development, based upon the characterization of the product, relationship with 

 behavior (when possible), or new findings. Table I defines the quality attributes, which tests are used to check the fulfillment of these quality attributes, and whether a quality attribute is critical for the safety, efficacy, and usability for the patient. The last column defines whether the CQAs are controlled by the process, by the materials, or by GMP systems. If the control relies on GMP systems, criticality management of the attribute is not pursued. 

Table I: Quality attributes of a drug product (tablet). 

The next steps—criticality in product, materials, and package—are similar to those in the manufacturing-process criticality management, which is described below. 

Materials play a role both in the drug product and in the manufacturing process. Therefore, they can be part of either the drug-product criticality management or the manufacturing-process criticality management. 

Manufacturing-process criticality management

Figure 2 outlines the steps in the manufacturing-process criticality management. The list of end-product CQAs and a detailed process map are the starting points. 

Figure 2 (ALL FIGURES ARE COURTESY OF THE AUTHORS.) 

In the first step, a team of experts identifies factors that might influence these CQAs and which process steps, process parameters, material attributes, and attributes of the pharmaceutical intermediate or in-process product are worthwhile to investigate further. This evaluation is based on prior knowledge and thus is a science-based risk assessment. This assessment is represented in a tabular format that provides a clear overview of the parameters and attributes that will be investigated further (see Table II). 

Table II: Potentially influential steps and process parameters or material attributes for a direct-compression tablet. 

Based on the overview, multivariate experiments are designed in a reasonably broad space. Experiments are preferably designed by DOE, where it helps and is supported by PAT (where it helps and is available), and stability studies, where relevant. 

Table III: Compression-process parameters and active pharmaceutical ingredient (API) attributes that affect critical-quality attributes or manufacturability. 

In the second step, the outcome of these experiments is used to define the process parameters, material attributes, and the attributes of the pharmaceutical intermediate or in-process product that influence the end-product CQAs. This evaluation is based either on statistical significance in the experiments or on strong prior knowledge. In the absence of statistical significance, the parameter or attribute is not influential and therefore considered not critical. Reasons and references to the experiments or reports are captured in a tabular format (see Table III). In this way, the criticality-management document represents the product knowledge and process understanding in a clear and concise way and can be retrieved quickly, with references to underlying results and details. 

Table IV: Target, normal operating ranges, design space, and risk analysis for parameters of a compression process. 

In the third step, for every process parameter or material attribute that proved to be influential, one must define the normal operating range and its part in the design space (see Table IV). The normal operating range and design space must take into account the multivariate combinations and interactions and be valid at full scale. If the manufacturing equipment has no advanced controls such as feed-forward- and feedback-control mechanisms, one must set a clear operating range for each influential parameter or material attribute. For example, one can reduce the operating space to a regular square or cube to allow the operators to understand the area in which the process must be controlled (see Figure 3).  

Figure 3 (ALL FIGURES ARE COURTESY OF THE AUTHORS.) 

When the design space is irregular, a mathematical equation can be provided together with the boundaries in which this equation is valid. If only two process parameters play a role, they can be represented by an overlaid contour plot. If more parameters play a role, then more or more complex graphs are needed to represent the design space. When the design space is large, one can simplify the representation of the design space by presenting it as a cube or square within the contours of the design space. Thus the design space is reduced, but it can be represented as a set of multivariate proven acceptable ranges: one for each influential parameter or material attribute, taking into account all combinations and interactions with other parameters or material attributes. Multivariate proven acceptable ranges defined within the design space are a substitute for the design space, but univariately defined proven acceptable ranges do not constitute a design space. 

Steps 1–3 in Figure 2 represent the knowledge part of criticality management. Steps 4 and 5 represent the risk management based on this process knowledge and the translation into a suitable control strategy. 

Risk analysis to define the intrinsic criticality of process parameters or material attributes

In the fourth step of criticality management (see Figure 2), one evaluates the intrinsic criticality of each influential process para-meter or material attribute. A risk analysis is performed to check and control whether the variation of influential parameters or material attributes can jeopardize an end product's CQA, stability, or manufacturability. The risk analysis is performed for each individual influential parameter or material attribute to determine its effect on individual end-product CQAs while still taking into account the interactions with other parameters or material attributes. Figure 4 shows the overall risk process. 

Figure 4 (ALL FIGURES ARE COURTESY OF THE AUTHORS.) 

There has been much debate about the definition of critical parameters and attributes. Not every influential parameter or attribute should be called critical. A distinction should be made between the vitally important process parameters or material attributes and those with minor, though statistically significant effects. 

A critical parameter or attribute has a strong relationship with the CQAs of the end product. Furthermore, the quality of the end product is likely to be affected if the variability of this process parameter or material attribute is not tightly controlled. 

The starting point in the risk analysis is that the product quality and its manufacturability are considered acceptable when the process parameters operate within the design space. If they operate outside the design space, the quality or manufacturability might be affected. However, the design space is not necessarily set at the edge of failure. One must determine what the risk is that an individual process parameter or material attribute will drift outside the design space and lead to a defective product if no specific controls are applied. The higher this risk, the higher the intrinsic criticality of the process parameter or material attribute. This intrinsic criticality is defined based on severity and probability. When robust model equations are in place, quantitative risk approaches might be used. It is also possible to use qualitative approaches to identify the critical parameters or material attributes that need tight controls. The severity (i.e., high, medium, or low) defines the potential magnitude of the effect on the end product CQAs when a process parameter or material attribute moves outside its design space. The magnitude in effect has to do with the strength of the relationship between the process parameter or material attribute and the end-product CQA—or with the closeness of the design space to the edge of failure. Probability (i.e., high, medium, or low) defines the likelihood that the process parameter or material attribute will move outside its design space when no special controls are applied. This likelihood has to do with the mean and variation of the process parameter or material attribute 

 the design space or with the location of the normal operating ranges 

 the design space. Probability should be rated without the controls to accurately evaluate the intrinsic risk in a process parameter or material attribute. Reference tables can be used to define 

 Based on the severity and probability scores, an intrinsic criticality rating (1 to 5) is attributed to each influential process parameter or material attribute (see Figure 5) and filled out in the parameter table (see Table IV). 

Figure 5 (ALL FIGURES ARE COURTESY OF THE AUTHORS.) 

Process parameters or material attributes with a high intrinsic criticality are designated as critical and need tighter controls such as higher sampling frequency and advanced control systems. It should be noted that the detectability score of an FMEA should not yet be used to define the intrinsic criticality because otherwise no parameter or material attribute will be called critical if good controls are in place. Indeed, a control strategy does not make a process parameter or material attribute less critical, it simply controls the parameter or attribute. 

Definition of a comprehensive control strategy and estimation of the residual risk

Based on the detailed product, process, and material understanding and the intrinsic criticality of parameters or material attributes, the development team should discuss and agree on a comprehensive control strategy. The strategy determines where in the process to install which controls to consistently guarantee quality (see Step 5, Figure 2). 

The more critical the parameter or material attribute, the tighter the controls that are needed. For example, the sponsor would of necessity place tighter controls on parameters that have the largest influence in the formation of a degradant. 

One may consider moving to advanced control systems such as PAT when necessary to guarantee product safety and efficacy and appropriate detection systems are available (e.g., during direct compression of low-dose tablets for drugs with a narrow therapeutic range). 

The control strategy defines measurement equipment; methods; sampling frequency; and size, target, and normal operating ranges for a combination of the following:  

Influential raw-material attributes (i.e., incoming control) 

In-process controls and PAT on influential quality attributes of the pharmaceutical intermediates or in-process product

Quality-control testing on the CQA of the end product. 

To rate how well the process parameter or material attribute is controlled, the team assigns a detectability rating in the parameter table (see Table IV) based on the matrix in Table V. Several aspects of detectability are important such as the timing (one must detect and act on a cause or an event before an effect occurs), the relevance of measurement (one must measure the right attribute or parameter), the quality of the measurement method, and the sampling frequency. 

Table V: Detectability ratings for process parameters and material attributes. 

Subsequently, the residual risk is estimated and evaluated. The main risk question to be addressed is whether the current level of control is adequate for the criticality of the process parameter or attribute. The team will attribute the residual risk level (i.e., acceptable or unacceptable) to each influential parameter or material attribute based on intrinsic criticality and detectability (see Table VI). The risk evaluation primarily focuses on the patient. In certain cases, however, risk controls might also be needed to lower the manufacturer's risk of discarding batches. 

Table VI: Decision matrix to determine acceptability of the residual risk. 

Articles

Criticality management combines pharmaceutical product, process, and material knowledge and risk management in one approach, which is reflected in a single document.