Quality By Design

A seamless connection between the laboratory, process development, and manufacturing

QbD relies heavily on scientific and engineering statistical methods, including design of experiments (DOE) and risk assessment techniques. It is based on the concept that quality cannot be tested into products but should be built in by design. QbD has its origins in product development and manufacturing. However, it has significant benefits for the laboratory. QbD can enhance the efficacy, robustness (tolerance to small changes in operating conditions), and ruggedness (sample test reproducibility for different standard test conditions—different analysts or instruments) of laboratory methods. QbD product development and manufacturing processes are highly dependent on optimized, reproducible, and accurate laboratory methods. QbD processes will not yield maximum benefits to industry and customers without a strong QbD presence in the laboratory.

QbD has received most attention in pharmaceutical product development and manufacturing, to a large extent because QbD can significantly improve both business and regulatory models. On the business side, QbD improves product design, decreases manufacturing problems, and results in fewer manufacturing supplements for post-approval changes and less regulatory scrutiny for new technology implementation. It also reduces cost, waste, and deficiencies and yields faster approvals. From a technical standpoint, QbD results in enhanced product and process understanding, including the interactions among ingredients and process conditions. This understanding yields higher-quality, more effective, and safer products.

On the regulatory side, QbD improves coordination in review, compliance, and inspection and results in more useful information in regulatory submissions. This leads to more consistency, flexibility, and improved review quality. These improvements benefit regulators who must review large amounts of data and documentation. QbD ensures that science is integrated in multidisciplinary decision making, which is more efficient for regulators than relying on empirical information alone. Finally, QbD allows for the allocation of resources according to risk, and that is good for business, regulators, and, most important, customers.^6-7

*Table 1: Manufacturing Process and Laboratory QbD Phases.*
Action	Manufacting Process QbD	Laboratory QbD
Define	Target Product Profile	Target Method Performance
Establish	Critical Quality Attributes (pCQAs)	Method Critical Quality Attributes (mCQAs)
Conduct	Risk Assessment	Risk Assessment
Develop and Verify	Design Space	Method Design Space
Implement	Control Strategy	Control Strategy
Conduct	Continuous Improvement	Continuous Improvement

QbD is a scientific, risk-based, holistic approach that results in product conception to commercialization by design. It is helpful to compare standard QbD for product development and manufacturing to QbD for laboratory operations. The most important QbD contribution in processing and in the laboratory is that analyses and resulting decisions are based on statistical data analysis and confidence levels and not simply on the analysis of specific sets of empirical information. Table 1 is a side-by-side comparison of process/manufacturing QbD and QbD for the laboratory. QbD starts with defining a target profile. The critical quality attributes that impact and generate the desired profile are then identified. Risk analysis, identification of the design space, control, and continuous improvement complete the process. In the case of process QbD, this framework allows for concurrent process and product design and development and manufacturing optimization to obtain a predetermined product quality. The analytical laboratory plays a key role throughout all phases of QbD, and it is an essential partner in the control strategy and continuous improvement of a QbD manufacturing process. QbD components with examples of implementation for laboratory methods are detailed below.

Target Product Profile (TPP)

This summarizes the features of an intended product. It provides the structure for an efficient product development program and includes all relevant technical and scientific information necessary for product development and commercialization.

Target Method Performance (TMP)

These criteria must be met by a method. Method performance criteria are derived from product-critical quality attributes (CQAs), as defined below, and must be clearly specified and understood. The required precision and sensitivity for a method are established from process specifications and tolerances. Allowable method variability must be specified, and the method must be sufficiently sensitive to the specification limits. It is also important to clearly identify variables to be measured and monitored and the selectivity required among various measurements (for example, impurities). The routine use of laboratory methods is part of TMP. Understanding process decision requirements is critical to the effective routine use of laboratory methods in manufacturing environments. Method precision, sensitivity and selectivity, and intended use are key elements of target method performance.

Process-Critical Quality Attributes (pCQAs) and Method-Critical Quality Attributes (mCQAs)

Figure 1: General Fishbone Cause-and-Effect Diagram for a Chromatography Method.These attributes potentially affect the efficacy and safety of the product or method. Cause-and-effect diagrams (fishbone or Ishikawa)^8-9 are widely used to identify pCQAs and mCQAs. Figure 1 illustrates a fishbone diagram for a chromatography method. In the figure, the cause categories are listed as measurement, personnel, environment, equipment, method, and materials. Each variable listed can impact the critical quality attributes of the measurement. Variables most likely to directly impact results are listed on the primary branches, as shown in Figure 1. Variables with secondary impact can be listed on secondary branches ancillary to primary branches. It is also possible to further classify variables as factors to be controlled (C), noise (N), or experimental parameters (X), the acceptable ranges of which need to be determined. The CNX classification helps focus experimentation. Understanding of CQAs evolves as the process or laboratory method is developed and during manufacturing or routine laboratory method use. Understanding of CQAs early in process or laboratory method development has many benefits. Design of experiments, typically employed in the determination of the design space, as highlighted below, can provide guidance in the final steps of CQA identification. Screening experimental designs such as Plackett-Burman¹⁰ can follow a fishbone analysis to identify mCQAs and pCQAs. As part of initial studies, screening designs can be used to pinpoint statistically significant factors from those identified through a fishbone analysis with a minimum of experimentation.

Risk assessment

This is a quantitative analysis of risk associated with products, processes, and laboratory methods. Failure mode effect analysis (FMEA)^6-8 is widely used to quantify risk in processes and in the laboratory. FMEA includes a list of all possible failures and their consequences. Risk for each failure is quantified by a risk priority number, which is the product of the assessed probability of failure occurrence, failure severity (impact on product efficacy and safety), and the detectability of the failure. Assigning values to these three factors is a team exercise and involves some qualitative judgment. A very important part of FMEA is specification of corrective actions for each possible failure. The FMEA team should include members with significant experience with the operation or product under review or similar operations or products. Knowledge bases and databases for processes and laboratory methods can enhance the efficacy of the FMEA team and improve analysis and resulting decisions. Priority matrices can also be helpful risk assessment tools.

Design space

This range of process inputs helps ensure the output of desired product quality. The design space establishes process ranges for variables. Design of experiments (DOE) is often used to identify the design space.^6,8,10 Processes then operate within the design space. DOE consists of a wide range of techniques to investigate process conditions that yield product of the desired efficacy and safety. DOE techniques involve changing more than one variable at a time. Figures 2a and b illustrate the difference between experimentation varying one variable at a time (one factor at a time: OFAAT) and DOE for a two-variable experiment. Figure 2a shows six experimental conditions holding pressure constant (P1 and P2) and varying temperature. Figure 2b depicts a two-variable DOE factorial design.

Figure 2: Two-Variable Experiment—(a) One Factor at a Time, (b) Two-Level Factorial with Center Point.Experiments conducted at the corner points (circles on Figure 2b) and at the center compose a two-factor, two-level factorial design (22) with a center point. Experimental designs have been developed for every phase of experimentation, from the screening designs discussed above, useful in identifying significant variables from a large number of potential considerations, to highly optimized designs for complex, highly nonlinear systems. Current USFDA initiatives are designed with significantly less regulatory oversight for process changes made within a QbD design space.³ Thus, in addition to optimizing processes and products, QbD has the potential to optimize regulatory operations for industry as well as for regulators. Among the many benefits of DOE are increased efficiency in experimentation (fewer experiments), the simultaneous and clear identification of variable effects and their interactions on the output, the analysis-ready data that allows for better conclusions, and accurate models for confidence levels established by process requirements.

Method design space

Method design space can be developed using DOE methods, as is the case for processes. In the laboratory, it is simpler to separate nuisance (noise) factors from method variables. Noise factors can be minimized independently using robustness analysis by varying noise factors to elucidate potential problems.

Control strategy and continuous improvement

Process monitoring, meeting target product profiles and process specifications, and introduction of technology and process improvements are critical to process safety, efficacy, and profitability. Analytical methods are key elements throughout the QbD process and are essential to the control strategy and continuous improvement phases. A significant process analytical technology (PAT) component is often part of QbD manufacturing processes. Sensors and measurements monitoring continuous or semi-continuous processes allow for real-time or near real-time adjustments that can yield the desired product profile at every process stage. Similarly, control and continuous improvement must be part of laboratory methods and general laboratory operations. The QbD systematic approach makes control and continuous improvement routine and ensures that laboratory methods yield their TMP.

Figure 3: Laboratory Quality by Design (QbD) Map.Laboratory QbD, as depicted in Figure 3, can create a seamless connection between laboratory, process development, and manufacturing. This connection is the foundation for control, continuous improvement, and enhanced understanding of all CQAs. Laboratory QbD includes statistical tools and methodologies that optimize methods and develop method-specific databases and knowledge bases. Results obtained for specific methods can provide insight for other methods and products. QbD-developed methods lead to fewer out-of-specification results, shorter turnaround times, more robust measurements, and more efficient technology and knowledge transfer. Continuous improvement is a hallmark of QbD. The databases and knowledge bases result in a more productive, flexible, and innovative environment. In addition, the nature of QbD fosters multidisciplinary collaborations, engendering a team approach to problem solving that enhances links between laboratory work, process/product development, and manufacturing.

QbD in the analytical laboratory can optimize laboratory operations and support 21st-century manufacturing. It requires a common language and industrial commitment to provide the necessary training for laboratory personnel to develop QbD skills. QbD methodology is not simply analytical technology transfer and ICH validation. It is a risk assessment–based scientific approach that results in method improvements from internal controls and method changes for analyses leading to improved target method performance and efficiency with regulatory flexibility and simplification.

References:

1. “US Food and Drug Administration, Pharmaceutical CGMPs for the 21st Century—A Risk-Based Approach,” 2004.

2. "USP 29-NF 24 United States Pharmacopoeial Convention," USP, 2006.

3. “International Conference on Harmonisation of Technical Requirements for Registration of Pharamaceuticals for Human Use, Quality Guideline Q8 Pharmaceutical Development,” 2006.

4. “The International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use, Quality Guideline Q9 Quality Risk Assessment,” 2006.

5. “The International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use, Quality Guideline Q2 (R1), Validation of Analytical Procedures: Text and Methodology,” 2005.

6. Z. O. Gephardt, “Quality by Design (QbD): A New Paradigm for the Analytical Laboratory I—Fundamentals and II—Experimental Design” Eastern Analytical Symposium Short Course, Somerset, NJ, 2013.

7. P. Borman, M. Chatfield, P. Nethercote, D. Thompson, and K. Truman, “The Application of Quality by Design to Analytical Methods,” Pharmaceutical Technology, vol. 31, no. 10, 2007.

8. R. Amurag and R. Mhatre, Quality by Design for Biopharmaceuticals, Hoboken, NJ: John Wiley & Sons, 2009.

9. S. Orlandini, S. Pinzaute, and S. Furlanetto, "Application of Quality by Design to the Development of Analytical Separation Methods," Analytical and Bioanalytical Chemistry, no. 405, pp. 443-450, 2013.

10. G. Box, J. Hunter, and W. Hunter, Statistics for Experimenters: Design, Innovation and Discovery, Hoboken, NJ: John Wiley & Sons, 2005.