Approach 2: Full design of experiments
An enhanced QbD approach to product development would additionally include a systematic evaluation, understanding, and refining
of the formulation and manufacturing process, including (3):
- Identifying–using prior knowledge, experimentation, and risk assessment–the material attributes and process parameters that
can have an effect on product CQAs
- Determining the functional relationships that link material attributes and process parameters to product CQAs
- Using the enhanced process understanding in combination with quality risk management to establish an appropriate control strategy
that can, for example, include a proposal for design space(s) and/or real-time release.
Risk-assessment tools can be used to identify and rank potential parameters deemed to have an impact on product quality based
on prior knowledge and initial experimental data. The initial list of all possible parameters can be quite extensive, but
is likely to be narrowed, as process understanding is increased, to a smaller list of potential parameters. Narrowing the
list has the advantage of reducing the number of experiments necessary in the modeling of a design space. The list can be
refined further through screening experimentation to determine the significance of individual parameters and potential interactions.
Once the significant parameters are identified, they can be further studied (e.g., through a combination of design of experiments,
mathematical models, or studies that lead to mechanistic understanding) to achieve a higher level of process understanding.
Figure 1: An Ishikawa (fishbone) diagram that identifies all potential parameters that can have an impact on the desired quality
attribute. (ALL FIGURES COURTESY OF THE AUTHORS)
Example 2: Full design of experiments.
Using prior scientific knowledge, the project team experts will use the target product profile to establish CQAs, and, in
turn, CPPs. An approach could follow the five steps listed below.
Table I: The potential parameters in an example analysis.
Step 1: Cause and effect analysis (Ishikawa diagram).
The expert team compiles a cause and effect (C&E) diagram, also known as an Ishikawa diagram, that maps out all potential
parameters in a manufacturing process. The number of parameters can be very extensive. An example analysis performed for a
recent project at Almac resulted in 79 parameters, possibly influencing the final tablet quality (see Figure 1). The parameters
determined are shown in Table I. Figure 2 shows the potential CPPs making up the dry granulation step of the manufacturing
Figure 2: Example C&E diagram detailing dry-mixing and granulation parameters.
Step 2: Ranking of process parameters in order of importance.
The list was analyzed using a risk-assessment approach to prioritize the parameters. Parameters graded as high risk were progressed
for investigation. An example parameter rating form is presented in Figure 3. The example shows a parameter rating form focusing
on the nine parameters from the dry-granulation stage that may have an effect on the CQAs. A separate rating form would be
completed for each stage of the manufacturing process. A scoring system, ranging from 1 being minimal effect, 3 being moderate,
and 9 having a major effect on product attributes, was used. This scoring approach applies a much greater weighting to parameters
deemed to impact a product's attributes significantly, thus increasing the necessity to further study the effect of the particular
Figure 3: Example parameter rating form.
Figure 4 shows an example of results following the compilation of the scoring process. The example illustrated focuses on
a single product attribute (tablet hardness) and a single processing step (dry granulation). The full exercise produces individual
tables compiling project team views regarding each processing parameter in relation to each CQA.
Figure 4: Example results parameter rating form.
Figure 5 shows how decisions can be made on which parameters may be investigated at the next stage. The total score is the
sum of the scores obtained for process parameters across all the CQAs. This will provide a rank order of parameters that may
have an effect on CQAs. Using prior experience coupled with knowledge of the intended process, the number of parameters can
be minimized. The number of parameters investigated using experimental design will significantly increase the number of batches
Figure 5: Parameter evalution to determine which parameters should be investigated.
Step 3: Assessing the impact of parameters on the product.
The parameters were further screened by considering the higher ranked parameters in each processing step as having a high
influence on each CQA. Also, every parameter having a score of 40 or more was classified as having a high influence on CQAs.
An example matrix showing the impact assessment for the dry granulation step is presented in Figure 6.
Figure 6: Detail of parameter risk assessment.
Step 4: Determine the parameters to be investigated experimentally.
Following a full analysis using prior knowledge and experience of the formulation and the manufacturing techniques, conclusions
will be drawn from the risk-based classification process to list the parameters to be investigated using a first-screening
experimental design (see Table II).
Table II: Parameters to be investigated in example analysis.
The parameters were investigated using a 2-level screening design of experiments. In this example, 10 parameters were chosen
to be screened.
The number of parameters, in addition to the desired resolution of the design, impacts on the number of experiments required
in determining the critical parameters. Figure 7 shows the relationship between number of parameters and resolution with number
of experiments required. Additional considerations such as time and active pharmaceutical ingredient constraints may also
influence the number of experiments that are possible at this stage.
Figure 7: The relationship between the number of parameters and resolution of design to the number of experiments required.
The standard experimental design notation is presented in Figure 8, where X is the number of levels being investigated. A
level is a particular parameter value. In the case of compression speed, the two levels could be the maximum and minimum turret
speed that would potentially be used in an operation. K is the number of parameters being investigated. P is the degree of
fractioning that is being performed in order to reduce the number of experiments. A full factorial design would not have a
value for p displayed.
Figure 8: Standard experimental design notation.
For example, 25 would be a two-level design covering 5 parameters. This would require 32 experiments to investigate the parameter effects.
25-1 is a fractional design of this example where the number of experiments required is halved to 16. As the degree of fractioning
increases the quality of information obtained from the design decreases. A typical approach would be to use a high-quality
(high resolution) design for optimization and in-depth study of a process and use a low-resolution design for initial screening
in order to determine critical parameters. The design would then generate a number of experiments that detail the exact parameters
to set for a manufacturing operation. Each experiment would be completed and the subsequent product attributes would be determined.
Statistical software is used to analyze the effect of the process parameters on the product attributes and determine how critical
each parameter is.
In the described example, the results of the experimental design allowed the number of potential critical parameters for the
initial screening to be reduced to eight. These eight parameters were then investigated using a more comprehensive 3-level
design. This type of design, also referred to as a response surface method (RSM), will require more experiments. The previous
2-level screening design is used to minimize the number of parameters investigated so as to minimize the number of experiments
at the 3-level design stage. In this example, the results of the 3-level design enabled statistically significant models to
be calculated showing the effects of the critical parameters on a number of product attributes. A graphical example of a model
displaying the influence of granulator final screen size and roller pressure on tablet hardness is shown in Figure 9. As can
be seen, as the roller pressure and granulating screen size decreases, the tablet hardness increases.
Figure 9: The influence of granulator final screen size and roller pressure on tablet hardness.
Step 5: Design space.
The linkage between the process inputs (input parameters and process parameters) and the CQAs can be described in the design
space. The risk-assessment and process-development experiments described can not only lead to an understanding of the linkage
and effect of process inputs on product CQAs, but also help identify the parameters and their ranges within which consistent
quality product can be achieved.