Several years ago I led a manufacturing team that was tasked with moving protein production processes out of development and into manufacturing.
Over the course of six years we moved more than 20 different products from IND to BLA stage. These came from various development groups and were at various levels of characterization. As a manufacturing team we decided early in this series that we needed to become very good at this, as there would be an advantage to our company and to our team to reduce the time for being ready to file the CMC portion of the BLA.
We successfully met this challenge: while the first process took us two years to ready the CMC information for the license application, by the time we reached protein number 20, we had reduced the time to six months.
This activity taught us many lessons. The first lesson was that technology transfer is hard (see Wheelwright 1994). Improvement in technology transfer skills requires serious and sustained effort and is not a trivial activity. The second lesson was that whenever we transfer a process from one group to another, it is difficult, if not impossible, to transfer everything. In particular, it is rarely possible to transfer the exact environment, and for the poorly characterized process the environment can be very important.
An example illustrates this point. Our typical start of technology transfer for the downstream part of the process began with a two-page protocol from development, which listed all reagents, solutions, chromatography conditions and so forth. Our initial goal was to duplicate the process exactly to confirm that critical details were listed in the protocol and did not reside only in the development scientist’s mind. In this case, when we in manufacturing ran the process in our transfer lab, we did not recover any protein. Upon communicating this result to the development scientist, she ran the process again in her lab with complete success.
However, when she ran the process in our lab with our reagents, she obtained the same result we did. After investigation we confirmed that our buffer solutions, which were prepared by a central service group, had wider allowable specifications than those to which the scientist prepared her own solutions. For manufacturing, our range was set by the need at the large scale to prepare our buffers by weight, not by volume, whereas the scientist used graduated cylinders and volumetric flasks.
In other words, the process as transferred was not sufficiently robust to succeed in a manufacturing environment. The solution was for the scientist to evaluate the optimization of the chromatography step and confirm the maximum acceptable range of preparation of these buffers. Once this was accomplished we in manufacturing had no difficulty in always delivering the purified protein.
To summarize this example, process transfer often includes a change in the environment (in this case, how buffers were prepared) that tests the limits of the process operating parameters.
To reduce this challenge in future transfers, our manufacturing team hired a process scientist whose primary job was to characterize the process at the start of transfer. We had unsuccessfully lobbied the development teams to provide this process characterization history to us, but recognizing the importance to the success of bringing these products to market, we paid for it out of our budget rather than do without.
Today most companies recognize the need for process characterization and collect this information as part of development, typically including it in development reports and making it available to the manufacturing team at the time of technology transfer .
While process characterization may be driven in many organizations by regulatory requirements, it is good manufacturing practice from an economic standpoint. There are times during the course of production when problems arise during manufacturing. The question of whether material that is produced under a process deviation may be sold in the market place may hinge on the level of experience we have with the process and the performance we have seen in the product under various parameter ranges.
Quality by Design (QbD) is supported by regulatory authorities because it adds to the confidence of the authorities in our capability to deliver products with consistent quality. There are several aspects of QbD that we should consider as we move our process from the laboratory to the commercial plant.
One of the first steps in understanding our process is identification of critical quality attributes (CQA) and critical process parameters (CPP) that have clear impact on the performance of our product in the clinic. Identification follows a conscientious analysis of our process, our test methods, and our product for risks that may lead to non-standard outcomes. A comprehensive risk analysis is necessary for us to identify all of the issues. A comprehensive analysis also enables us to apply a somewhat quantitative assessment to each issue, enabling us to rank both the probability of occurrence and potential consequences.
Along with understanding CQA and CPP, we need to ensure our process is optimized with respect to robustness. A robust process is one that gives the same output (as measured in quality and quantity) even when the inputs vary. In the case of biological processes, there is always some degree of variation in the upstream process because we are dealing with living organisms. In most processes there is variation in raw materials, within some acceptable range as delineated in our raw material specifications.
The goal of a robust process is to ensure that regardless of the variations that occur during the process and in our starting materials, the product we make is always safe for the patient and within product specifications so that we maximize the return on our investment (and avoid rejected batches) .
QbD is what we call the formal methodology for developing our robust process. There are many tools that are useful to bring us to a robust process, including design of experiments, statistical methods for data analysis, process control using process analytical technologies and statistical process control, root cause analysis, and analytical method validation.
Regulatory agencies now require us to employ QbD and include descriptions of our process design as part of our license application. This is not something we can satisfy with vague descriptions; we must employ a rigorous methodology, documenting our data in formal reports.
There are many literature sources that provide detailed information on how to apply QbD to our process optimization and method development. We can employ these resources and the assistance of specialists to ensure we meet the expectations of our regulatory authorities .
But hopefully we see that beyond just following the minimum requirements, there are benefits to our organizations from development of robust production processes. The benefits that accrue from process characterization and robust process development include enhanced patient safety, fewer failed batches, and reduced manufacturing issues.
References
1.Wheelwright, S.M., "Commercializing Biotech Products," Bio/ Technology 12:877-880 (1994).
For additional detail see www.qualitybydesignconsulting.com
Scott M. Wheelwright, PhD is founder of Complya Asia, a consulting firm in China that provides support to companies sourcing from Asia (including vendor audits for GMP and GLP compliance) and project management support to ensure vendors meet quality requirements and delivery schedules. Complya Asia also provides support to companies in Asia that seek to meet international standards for GMP compliance .