Sue Pearson Ph.D. Freelance Writer GEN
Luxury or necessity for developing the best biologicals?
“There is the perception in biopharm companies that it is too expensive and takes too much time to implement full QbD (Quality by Design) for biologicals,” states Professor Jose Menezes, Ph.D., Institute of Biotechnology and Bioengineering, Portugal. Dr. Menezes speaking at the recent PAT and QbD Forum 2014 at the Sartorius College in Goettingen, Germany adds, “Without considering the benefits over a product’s lifecycle the cost would hardly be justifiable. However, if you take into account the product’s lifecycle, you’ll see increased process and product knowledge which will mean less batches will be lost, and a more reliable process that is less likely to have quality and recall issues so in the longer term your drug substance/product will actually be less expensive to produce.”
Mario Becker, director, PAT & automation at Sartorius Stedim Biotech adds, “Key factors driving the implementation of PAT and QbD in the biopharmaceutical industry are economic, particularly those concerning the optimization of both new and existing processes to deliver higher yields and titers, and time to market. These involve accelerating process development times and lowering the associated R and D costs.”
QbD in the biopharm industry is a systematic approach to developing a drug or vaccine and its key elements are defining a target product profile (TPP), identifying the critical quality attributes (CQAs), and then defining a manufacturing process to ensure the CQAs meet the TPP. This all sounds quite logical and straightforward but as many of the speakers at the forum noted it is often anything but this.
Model and Measure
“With QbD of biologicals you have to do a risk analysis to identify by what range you can vary upstream and downstream bioprocess conditions without directly affecting the CQA,” Anurag Rathore, Ph.D., professor in the department of chemical engineering at the Indian Institute of Technology Delhi, explains. “Then you can do your Design of Experiments (DoE) but for a full factorial run of 10 factors that might affect CQA you are looking at running 1,000 experiments and often there are more than 10 factors that can affect quality. For most companies this is just not feasible and is where using high-throughput process development models and multivariate data analysis (MVDA) on a continuous basis is essential.”
Petter Moree, director, global product management at MKS Umetrics agrees, adding, “Scientists can generate upstream DoE bioprocess run information using an automated microbioreactors system, and this can be plugged into their MVDA analysis software to provide a process understanding to determine which parameters affect the CQA of their biologics. However, this does mean investing time in data analysis when it is collected rather than waiting until the QC people have analyzed it because by then they might have lost batches.” According to Moree, Lonza has presented implementations of scaledown models and real-time MVDA with its media development and by removing unwanted variation in one of their bioprocesses has increased the yield of a biological by 75%, decreased cycle time by 40%, and increased throughput by threefold.
One company that is embracing the real-time measurement philosophy in its upstream bioprocess development is GSK. Andrew Heinrich, senior scientist, biopharm process research states: “We are generating domain antibodies and have to find good scalable methods of measuring cell density online.” Heinrich presented a case study using a cytokine product expressed in CHO cells, which were cultured in shake flasks, 2 L and 50 L bioreactors. Cell density was measured using an Aber instruments biomass probe and offline using a ViCell cell counter and a Trypan blue assay. The Aber probe measures capacitance and because it scans every 30 seconds Heinrich’s team used SIMCA MVDA software to analyze the large amount of data it produces. “The Aber probe shows good correlation with the offline cell density measurements in 50 L and 2 L bioreactor studies,” says Heinrich. “Realistically it is not practical to use the Aber probe with 100 shake flasks but it can be used with benchtop bioreactors. We are now moving towards single-use technology much more and are using the ambr 250 automated microbioreactor workstation as a workhorse with the Aber probe for cell screening as this means we can use our online measurements as feedback control to increase our in-process knowledge.”
Full Fat QbD
For companies that want to perform a full QbD for biologics production, then according to Dr. Menezes they have to do an end-to-end analysis, which involves analyzing the raw material of their media batches and their cell banks, as well as the upstream and downstream bioprocess conditions, and doing this during development and at specific points in time during routine manufacturing (the product lifecycle). This can help predict the impact of defined sources of variability on their product quality and help improve consistency through increased process knowledge.
“QbD is helping us to realize that you have to take a systems’ view and that the current QbD formulation is suboptimal, if as expected the different systems’ components interact with each other,” Dr. Menezes says. “The current QbD formulation is not equivalent to the design-for-six-sigma approach that can really ensure an overall optimal solution in terms of product consistency over its lifecycle.”
This concept that putting together optimized stages may result in an unknown impact on CQA was demonstrated by Fai Poon, Ph.D., director of cell culture at Hisun Pharmaceuticals. “We develop biosimilars and in 2011 we wanted to optimize the titer of our first biosimilar so we worked with different feeds and found we could improve our titer by six fold,” Dr. Poon explains. “But by optimizing the culture media supplement we decreased the sialic acid content of our media by 20%, and this meant the antibody yield decreased because it lead to low yield purification so it had a major impact on the CQA of our biosimilar product.”
To overcome this problem, Dr. Poon and his team looked carefully at the media components with the aim of improving the sialic acid content by 20% while still retaining their titer increase. They analyze the components of four feeds all based on the commercial feeds: HyClone™ Cell Boost 5™ Supplement; Irvine Scientific BalanCD™ CHO Feed 3; Life Technologies CD EfficientFeed™ C AGT™ Nutrient Supplement and SAFC® EX-CELL® CD Hydrolysate Fusion. They then performed a Principle Component Analysis (PCA) using SIMCA MVDA software, and this identified six significant components that are critical to the sialic acid content. They then added the six components to their media and performed small-scale spin tube experiments with their antibody clones to determine which components would increase titer and sialic acid content. Two components increased sialic acid content by greater than 20%, and these were then used in scale-up work at 2 L, 30 L, and 1,500 L to show that the sialic acid and titer increases were reproducible at large scale. Dr. Poon concludes: “Titer improvement can sometimes lead to changes in CQA. This is where MVDA is useful to pinpoint which media components are important and is a potentially important media development tool.”
Data is King
“QbD is really all about knowledge management but in reality gaining access to prior knowledge about a molecule to define the TPP and CQA is very difficult,” says Andreas Schneider, vp/international business leader, custom biotech at Roche Diagnostics. “For example, clinical data is not always accessible as it sometimes proprietary to the country or company where the original development or clinical work has been performed. If you cannot find out if a factor is important or not to the CQA, this will affect how you set up your DoE for defining the manufacturing process.”
Schneider concludes: “Sometimes with a quantitative risk profile of a product you’ll have a moderate or significant ranking, but what you really need are quantitative values to define the cut-off ranges. There is a need to scientifically argue the quantitative ranges and to provide that data with the QbD filing documents, but this requires seamless access to a massive amount of data. At Roche, to ensure we can provide the knowledge management for QbD, we have implemented a project called DAMAS (Data Acquisition Management Analysis System). For DAMAS, we have integrated 200 devices that run and store data generated from 70,000 samples per year from different business units, so that we can really get an overview of the data from our analysis systems.”
According to Schneider, this integrated centralized data management concept is required as drugs will pass through the FDA filing processes significantly faster if they are linked to an FDA fast-track program. The evidence is that in 2013, the Roche drug Gazyva was launched as the first biologics in a FDA “breakthrough therapy” program, and it was filed with a full QbD approach including the design space. The timelines to pass through the clinical trial phases were significantly less compared to standard filing processes so that Roche and Genentech are going continue to file any future drugs that have breakthrough therapy status supported by data management and QbD principles. This type of resource intensive approach to develop knowledge management for QbD may be available in big pharma but, for smaller biotechs, what is possible in a QbD context?
Most small biotechs don’t have time to do a full QbD submission, but they do use QbD tools and automate their manufacturing processes where practical as QbD is really all about reducing variability. So in a successful biotech what does QbD lite look like? In general, the TPP and CQAs for protein-based therapeutics are defined around upstream bioprocess parameters. Staff performs a DoE using temperature, pH, and feed in shake flasks, 2 L bioreactors or automated microbioreactors to model their 500 L scale production batches. For bioreactor control they use automated pH and DO probes for offline sampling and for online Process Analytical Testing (PAT) they use mass spectrometers to measure gas production and biomass probes to determine cell mass. They also have alarmed remote controls on their bioreactors and use software to monitor the bioreactors remotely. Using automation takes out the element of allowing bioreactor operators “to go rogue” and sample when it is convenient rather than at set time points. Also because staff can monitor bioreactor runs at home, they can bring bioprocess runs on track without losing the batch if they need to.
Where Next for QbD?
Many speakers at the QbD forum agreed that the major driver of implementing a full QbD strategy is to gain a deeper process understanding and in doing so produce more robust and safer biologics. This means that the use of single-use facilities where automation and online measurement is much more prevalent is going to be the trend going forward. Becker states: “A major aspect is product safety, which is primarily affected by the robustness and flexibility of the processes developed. This is precisely where the combination of innovative, robust process analytics, model-predictive control, and flexible automation is going to play a central role.”
Amit Banerjee, Ph.D., research fellow global biologics at Pfizer says, “For process understanding of the cell culture we have to have good scaledown models, which show equivalence at 2 L, pilot scale, and 12,000 L.” Dr. Rathore adds: “I predict all companies developing biologicals will be using scaledown models and MVDA in the next decade.”
According to Schneider the biopharm industry is lagging way behind other industries and needs the right data management in place. “QbD has a very strong IT component, which means that data quality matters and the IT stakeholders have to be involved,” he says. “Everyone fears the IT guys but they shouldn’t; if they are involved at the beginning it may add delay, but it will speed up the process in the end if they are fully engaged in the project.”
“Full QbD starts from a comprehensive control strategy of your raw materials and cell banks,” Banerjee added. “This is time consuming to do and implement, and perhaps leads to the proverbial pot of gold in the end for those companies that can do it well.”