Quality by Design A-MAb Case Study Challenges Conventional Thinking
Now that the project has been completed, John Berridge, PhD, another one of the working group’s three facilitators, thinks they achieved that goal. “When people read it, they will find areas that they agree with, and almost certainly find other areas where they don’t,” he said. “It’s an aspirational document.”
The case study, which discusses a fictional monoclonal antibody product, referred to as “A-MAb,” was developed over the course of a year by participants of the working group representing Abbott, Amgen, Genentech, GlaxoSmithKline, Eli Lilly, MedImmune, and Pfizer.
Two examples of where the case study challenged conventional thinking, Berridge said, related to scale and change management.
The limits of the design space for the 15,000-L production bioreactor step were largely based on data derived using a 2-L scale-down model, built on extensive development and manufacturing data (at both the 2-L and 15,000-L scales) for a previous approved product, “X-MAb”.
To demonstrate that the 2-L scale-down model was representative and predictive of large-scale manufacturing performance, the team developed a principal component analysis (PCA) model. PCA transforms a large number of possibly correlated variables into a smaller number of uncorrelated variables, called principal components, and analyzes the variability in the data. The analysis included 13 variables, such as peak viable cell density, final viability, pH, glucose, lactate, and peak lactate, and a high degree of correlation was seen in the data sets for the different scales. Also, a complete comparability analysis for the product was made at the 500, 1,000-, 5,000- and 15,000-L scales, and an “engineering design space” was developed for the bioreactor design. The result, the report says, supported the use of the scale-down model.
Lifecycle Approach to Process Validation
The report also recommends a validation strategy that “relies more on the continuous process verification rather than a minimum number of ‘validation batches’ typically practised.” For the bioreactor operation, the validation strategy involved just two batches at the 15,000-L scale to confirm that the process performance at the 15,000-L scale was within the model predictions. These two batches are seen as “the start of the continuous process verification process” and a lifecycle approach to validation.
To provide continued assurance that the process would remain in a state of control throughout the life of commercial manufacturing, the team would create a multivariate statistical partial least squares (PLS) model, which would ensure that internal correlations among variables is also considered. “For example, if at any given time the titer is lower than expected for the measured variable cell concentration, the PCA model will be able to detect this as a potential out of normal signal, even if both parameters are within their respective univariate ranges,” the report says. In this way, the PCA model can detect a large number of potential shifts, trends, and excursions that would not be detected by univariate monitoring tools.
Scale-up From 15K to 25K Considered Within the Design Space
The report also anticipated that the A-MAb bioreactor process would be scaled up further, to the 25,000-L scale. For the case study, it was assumed that the 25,000-L plant had an extensive and proven commercial manufacturing record of cGMP compliance and MAb production. Based on bioreactor design and engineering parameter characterization, the 25,000-L bioreactors were considered to be within the engineering design space, thus providing a very high degree of assurance that operation at this scale will result in comparable process performance and product quality. “Thus, the scale-up to the 25,000-L bioreactor is considered a movement within the engineering design space,” the report says.
If a change to a different bioreactor (e.g., one with a different impeller design or geometry) were considered, the report says that an assessment would be made to determine if the bioreactor characteristics fell within the engineering design space. If they did not, then equipment modifications or changes in operational parameters would be considered to bring the bioreactor operation within the approved engineering design space.
Highlights and Workshops
The best way to read the 278-page report, Berridge says, is to start with the introduction and chapter 2, then move to the highlighted information that appears in blue boxes throughout the report. “If the information in a particular blue box is of interest, you can read the surrounding details,” he says, “If not, you can go to the next chapter.”
In addition, members of the working group will participate in various public workshops hosted by ISPE and CASSS, who have been designated as the primary facilitators of education related to the case study. Sessions will be held during three 2010 CASSS events, including the Well Characterized Biological Products (WCBP) conference in Washington, DC, in January, the CMC Strategy Forum Europe in Vienna, in April, and the CMC Strategy Forum in Bethesda, MD, in July. Workshops also will be incorporated into already scheduled ISPE 2010 meetings, in Milan in March, Tokyo in April, Washington, DC, in June, and Brussels in September.
The entire case study can be found on the CASSS (www.casss.org) and ISPE (www.ispe.org) web sites.
QbD Case Study Will Push Limits