Russell C. Marians, PhD - Associate Director of Analytical Development, Gates Biomanufacturing Facility
The latitude to optimize a manufacturing run in response to in-process test results is often omitted from cell therapy INDs in exchange for speed to clinic. My intent in writing this article is to convey the importance of utilizing in-process analytical testing in fixed processes, even at the earliest phase of clinical manufacturing, as that testing can directly affect patient outcomes. For this story to make sense, a little background is required. Put in place to support investigator-led Phase 1 clinical trials, the Gates Biomanufacturing Facility (GBF) opened its doors in 2015 to be the cell and biologics manufacturing arm of the University of Colorado Anschutz Medical Campus. This discourse will focus on the cell therapy branch of the organization, which had its first cGMP run in 2018. Since then, we have manufactured under five INDs, three of which are still active. Even though we are a relatively new facility, we can accommodate up to five runs a month and are quickly approaching our 100th cGMP cell therapy run, encompassing products from both university and private sponsors.
The concept of Process Analytical Technology (PAT) is not a new one: identify trends through a series of in-process analytical tests that are predictive of manufacturing failure and success. Prime examples of such tests are monitoring pH, glucose, lactate, and dissolved oxygen in both prokaryotic and eukaryotic fermentation. Advances in technology that miniaturize analytical instrumentation and facilitate in-line testing, coupled with streamlined test methods, expand the reach and utility of in-process testing to new and potentially more informative analytes. When anomalies in any given analyte level are recognized in real time, a process can be adjusted in order to prevent deviations and out-of-specification batches which results in reduced costs and increases product efficacy and safety. However, balance and discipline are required to combat the desire to collect as much data as possible, a common trait in most scientists. Generating actionable data is a cornerstone of any PAT philosophy, so an established, well-understood process with clearly defined and validated critical quality attributes (CQAs) are needed to enable on-the-fly process changes. But wait, this story is from the perspective of a Phase 1 site.
Phase 1 cGMP is a lot of things, but well-understood, clearly defined, and validated are not usually amongst them, especially in cell therapy. Assuming data from three full-scale runs were included in an IND, it likely means there were, perhaps, four full scale runs performed. This isn’t because cell therapy teams are data averse, but instead that cell therapy processes are expensive and resource intensive (both in time and personnel). Those issues are compounded if viral transduction is used to introduce a therapeutic gene into cells. The cost of viral vector used in one manufacturing campaign can run upwards of $5,000. With a lead time of 1-2 years and total investment of $2MM, it is no wonder such critical reagents are reserved for treating patients and not for development. To further complicate matters, most cell therapies are intended for autologous use. Cell therapy is the very epitome of personalized medicine. A patient donates critical raw materials (peripheral blood mononuclear cells in the case of CAR T cell products) used to create their own medicine. Genetic diversity paired with each patient’s journey through disease and treatment increases raw material variability right out of the gate. One manufacturing run leads to one dose for one patient. Although the field is actively working on allogenic therapies, where hundreds or even thousands of doses are produced from a single manufacturing run, that science has yet to come to fruition which means each cell therapy product still requires a unique manufacturing process. Commonalities present themselves after seeing enough products, but each product will present its own unique challenges and opportunities. That was a long way of saying there is little economy of scale in cell therapy. Not in materials, not in efficiency, and not in PAT.
Now let us consider the element of time and the challenge cell therapy products can pose to QC and QA teams. Cryopreserved product comes with the illusion of time and generally has a release window of 10-14 days. On the other hand, product that is infused fresh, within hours from the end of a manufacturing campaign, requires real-time testing and documentation review in order to preserve viability and function of the living drug. Both scenarios are in stark contrast to small molecules and traditional biologics that, generally speaking, can be archived and tested when convenient. Further complicating matters, the unknowns that accompany early phase manufacturing means samples for release testing can come out of the cGMP suite anytime within a 3-4-hour window, depending upon what new and exciting process challenges surfaced during that production run. Late samples that require immediate testing can further stress quality teams and cause unplanned shift work that creates a negative ripple effect by forcing delays to the following day’s docket as operators recover from an ad hoc late night of testing. Start with continuous improvement activities that occupy a quality unit, layer on product release for five cGMP runs per month, then toss in an unplanned shift or two and it is easy to see how quickly time can pass.
Collecting data about critical quality attributes (CQAs) is not new nor unique to cell therapy, but it is challenging in Phase 1 when a manufacturer is unfamiliar with the process and the Sponsor has little data supporting product efficacy. To that end, I suggest for consideration the best PAT readouts not only demonstrate manufacturing success but also can be used to predict clinical success. Correlates between PAT and clinical success are easier to draw by the time a therapeutic goes commercial but cannot be made in the earliest phases of manufacturing and clinical evaluation. However, unlike most experimental treatments, Phase 1 cell therapy trials are often conducted in a diseased population, making correlations between manufacturing data and clinical outcomes possible from the earliest clinical stage. Some manufacturing data may seem important to the manufacturing team but have little overall impact on efficacy. In that instance, should those data be considered a CQA? Perhaps. We can easily monitor the ratio of chimeric antigen receptor-expressing CD4+ and CD8+ T cells over any number of manufacturing runs, but what does that mean to overall manufacturing success, and more importantly, what does it mean for patient outcomes? At academic institutions, investigative therapies can not only help patients in the clinical trial, but also generate data throughout the study to inform the next generation of medicines and have payoff in the future. This forward-thinking mentality pairs with the desire to collect as much data as possible. Manufacturing data, especially when correlated to patient outcomes, are powerful hypothesis generators that can enable many grant submissions and further drug development efforts. The question then becomes which groups test and analyze in-process samples. There are many teams of skilled scientists at GBF capable of doing so, but generating research use only data can be distracting. The process development team does not always have the compliance mindset needed to operate in a GMP environment. The manufacturing team is best at manufacturing. QC should be laser focused on release testing. The common thread: All parties want to know the answer; all parties want to be helpful; all parties want to help the patient. Therein lies the answer. Data collection, tracking, and interpretation should be a team effort. It takes a village.
Yet another complexity is the sheer volume of data generated from a single cell therapy run. Take flow cytometry as an example. A typical panel that interrogates 10 different cellular proteins generates more than 100 different data plots, each telling a different part of the story, all of which need second scientist and QA review. So much data can easily overburden the quality teams and even the manufacturing team, when testing is performed in-line. Merely generating the data, although a required first step, cannot lead to action without organization. While it can be burdensome, building and maintaining data trackers and control charts requires a lot less time than doing the data dive for each out of trend or investigation. However, not all data are actionable. Facing large amounts of data without knowing which are important, mixed with the overwhelming desire to enhance process and patient outcomes can lead to another common pitfall, data over-interpretation. Part of the human condition is to fill-in knowledge gaps with our own pre-conceived notions which are always biased by tangential experiences and our own version of common sense. This is especially true in the case of Phase 1 clinical trials when the model is not well known. I suppose that is an argument to approach Phase 1 as solely a data gathering exercise, but that would leave money on the table. Deciding which bits of data correlate to the desired outcome and interpreting those data at face value are needed to focus data collection and process monitoring efforts as production continues.
We collect a lot of data on our Phase 1 cell therapy products, mostly using multiparameter flow cytometry focused on efficacy-indicating markers pulled from the literature. Unfortunately, those data don’t have much relevance to manufacturing success. Many of the usual PAT analytes, such as glucose and lactate, are not available to us because our processes are conducted as fed-batch rather than perfusion systems. For cell therapy developers, speed to clinic is often the driver, leaving little time for process development. To support that, there are several fed-batch systems on the market that are well-known to the industry. As a result, canned processes optimized by bioreactor vendors are incorporated into INDs, leaving no room for mid-process adjustments. However, that does not mean one should opt out of data collection. After all, the mantra still holds: you can only change what you measure. We try to focus on impactful data that support key release criteria and clinical outcomes. One attribute is very important to track: cell viability.
Cell therapies are living drugs. The link between cell viability and manufacturing success is obvious when live cells are needed to confer a therapeutic benefit. Consequently, there are go/no-go decisions around cell viability at no fewer than four different unit operations within a 10-day campaign for many of the products we manufacture. The test is easily performed both at-line and in the QC lab. One of those decision points comes right before transduction, which is the act of exposing cells to viral vectors loaded with the therapeutic gene. In an early run of a new product, we noted viability was well below the expected value. Not out of trend, mind you, because meaningful trends cannot be established with only 4-5 runs of experience, but suffice it to say the value was out of line with expectations. Knowing transduction would further decrease viability, all the relevant stakeholders (assay development, manufacturing, quality, regulatory, and clinical) were consulted and the decision was made to terminate the run prior to transduction. Afterwards, the clinical team had to make a tough phone call to the patient. During that call, it was discovered the patient had completed a series of radiation treatments just prior to donating the blood needed to initiate the manufacturing process. Manufacturing was reinitiated after the patient had sufficient time to recover from radiation, which led to a successful manufacturing run, a successful infusion, and, to date, a healthy patient. There were several wins in this example worth highlighting. First of all, the system worked. The appropriate CQA was monitored and identified a batch that was likely to be substandard. The finding was elevated before a critical reagent, the vector, was consumed on what surely would have been an out-of-specification run. Most importantly, the benefit in this case was very tangible in the most meaningful way possible. Not only was the patient treated, but the patient was treated about two weeks earlier than if the original campaign had been completed. Because that vial of vector was not used on an out-of-specification run, there was material enough to treat one additional patient. Potentially two lives enhanced by a single in-process cell count. This is why in-process testing is so important.
Author Biography
Russell Marians is the Associate Director of Analytical Development at the Gates Biomanufacturing Facility/ University of Colorado Anschutz Medical Campus. An industry professional for 17 years, Russell has worked in biologics development, device and diagnostics, and cell therapy. With four cell therapy INDs under his belt, Russell and his team help manage technology transfer, process development, and analytical development at GBF. Russell earned his BS in Biochemistry from New Mexico State University and his PhD in Biological Sciences from the Mount Sinai Graduate School of Biological Sciences of New York University
Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special
offers from American Pharmaceutical Review delivered to your inbox!
Sign up now!