7
Potential Next Steps to Consider for Addressing Variability
In the final session of the workshop, Carl Burke, Karen Christman, Allison Hubel, Peter Marks, and Clive Svendsen reconvened to expand on several of the topics addressed during the course of the day. The discussion, which included participant input, was moderated by Kathy Tsokas. Following the discussion, Tsokas summarized and shared ideas for moving the field forward with regard to the development of regenerative products based on presentations and discussions held throughout the day.
CROSSCUTTING TOPIC DISCUSSION
Getting the Most from the Available Data on Variability
Tsokas asked the panelists to comment on using the volumes of available data to inform discussions and decisions about variability. The quality of the data is a key factor in using available data, Svendsen said. Often one does not know the quality of data produced by another laboratory. This is why standards would be valuable. He also said there is a need for easier ways to share data, perhaps through ClinicalTrials.gov or the U.S. Food and Drug Administration. He cautioned that there would need to be some sort of data quality screening process. Burke agreed and raised the issue of variability versus uncertainty. When the data are not what was expected, it is not always clear if it is true variability or instead, for example, an issue with the assay. He raised concerns about researchers sharing data that they do not have confidence in. Hubel added that reproducibility is an important factor in developing confidence in the data being shared. Joseph Wu also
commented on the need to be aware of the quality of shared data. Referring to a biobank he discussed earlier (see Chapter 3), he said that the data generated and submitted by researchers receiving biobank specimens can be “messy.” He said, for example, that RNA sequencing data variability is often affected more by batch-to-batch variability than by biological variability. Normalizing such data is difficult.
Participants discussed sharing indication-specific or target-specific data that might inform the selection of clinical endpoints, for example, or the identification of critical attributes. It is possible that sharing data might help identify quality attributes for CAR T cell products, Marks said, but he added that different constructs can have different characteristics (e.g., adverse event profiles), complicating the interpretation of shared data.
Sharing can depend on the type of data, a workshop participant said, and there could be venues for sharing precompetitive data, such as data for standardizing analytical assays. However, he added, sharing clinical data for a proprietary product can be “a whole different question.” When clinical studies are published, the raw data are not necessarily included, Svendsen said, and it is generally the raw data that researchers want to see. He suggested that there is a need for a database where raw data from clinical trials could be uploaded for others to assess. “There is an enormously rich treasure of information already generated in the United States that is totally useless because it’s locked away in a vault,” he said. He suggested, for example, that those with expertise in artificial intelligence might be able to pull out insights from the data that the originators did not have the capacity to do. There might also be opportunities for precision medicine, for example, to correlate data with genomics. Svendsen added that companies do not necessarily want to keep their clinical data private.
Marks said that in his experience companies are not against data sharing; however, he added, a key element of data sharing is trust. A concern that is raised about sharing raw clinical data is that, in the absence of the associated statistical analysis plan, someone else can come up with very different conclusions about the endpoints. Marks mentioned several cases in which a re-analysis of raw clinical data by someone else drew an opposite conclusion about the effectiveness of the intervention being studied. He said that GlaxoSmithKline has a system designed to allow queries to the company’s data in a more organized manner. It comes back to being able to trust that those who have access to the raw data will not misuse it, Marks said.
Muschler discussed establishing a cell therapy registry. He said that not all of the parameters to be captured might yet be known, but data on the outcomes being measured can be standardized and shared. As a model, he mentioned the American Joint Replacement Registry, which collects and shares data on joint replacement outcomes. Marks responded that such a registry, “if done right,” could be very helpful. He said that there are several
groups discussing launching this type of registry. He emphasized that it is important to standardize the data that will be captured, such as endpoints, and ensure that the data going into the registry are from well-defined products. Muschler said that questions to be addressed would include who owns the data and who would be allowed to analyze them and publish them. Svendsen raised the topic of open source data sharing. The Allen Institute, he said, is an example of people working together to produce libraries of open source information.
Regarding data standards, another workshop participant proposed a mid-level of standards, perhaps “good sense protocols,” that would not be as rigorous as good manufacturing practice but that would standardize the collection of key data elements for the sample, such as the age, sex, and medical diagnosis of the patient.
Using Functional Assays to Assess Products
Participants offered a range of views and concerns regarding the use of functional assays to address variability of products. Siegel reiterated a theme from the morning discussion that addressing the variability of cell products would require moving beyond phenotypic assays to using functional assays. While this is conceptually true, he said that, pragmatically speaking, functional assays themselves add multiple dimensions of variability. He asked how realistic it would be to try to ensure the uniformity of a cell population by functional testing. Svendsen reiterated a comment made by Marks (see Chapter 6) that, in the face of multiple mechanisms of action, pick one assay and go with it. As an example, he said that in a case in which motor neurons were injected into the spinal cord, cell survival in the spinal cord was the key aspect to be assessed. He suggested that the critical aspect of whatever functional assay is chosen should be that it tests the reliability and consistency of the product in improving whatever it is being administered to improve. Even if the product is variable to a degree, a parameter related to the outcome of interest is the functional parameter to be assessed. Burke suggested that it is more difficult for cell therapies that do not have a clear mechanism of action and in which surrogate assays are used. But Svendsen said he felt that knowing the mechanism of action was not necessary if there was a functional assay. Marks said that a functional assay is even more important when the mechanism of action of the disease is unknown. Burke agreed that a functional assay is the end goal, but he added that, for an allogeneic treatment in very early development, functional assays can layer on more uncertainty. Hubel raised concerns about functional assays for products with multiple modes of action, and Siegel said that optimizing the “wrong” assay would affect the development of the product. Svendsen emphasized the need to start somewhere and said he
felt that it was important to develop some in vitro or in vivo outcome assay that reproducibly supports product outcomes in humans.
A workshop participant added that FDA is encouraging product sponsors to start thinking about multiplex potency assays early in development—a step that could help identify relevant biological assays. (See the FDA guidance document on potency tests for cellular therapy products.1)
Interdisciplinary Research, Collaboration, and Education
Panelists were asked to comment on interdisciplinary research and collaboration, not only across research and clinical disciplines but also including business and manufacturing sectors. Education and training are important elements for making interdisciplinary integration work, Hubel said. Each discipline “looks at the processing of the cells through a different prism,” she said, so common terminology and a common understanding of the critical issues are needed. Christman agreed with the need for education and training. She also noted the value of bringing some industry practices into the academic laboratory, such as developing standard operating procedures. Communication is important, Svendsen said, and clinicians, biologists, and engineers should come together in the same room. In many cases one discipline works for an extended time on a given technology, only to be told by the other discipline that it will not work. Communication and exchange of ideas needs to take place regularly.
Delivery of Care as a Source of Variability
Hubel emphasized the need to also consider the delivery of care as a source of variability. The location where the patient is treated with the regenerative medicine product is a variable. The number of sites for clinical trials is currently small, but as products reach the market, providers will handle multiple patient populations and multiple products, each with different administration and handling protocols. Providers have been asking for technology to help them deliver this care, she said, and providing that to them could reduce this aspect of variability.
FINAL THOUGHTS
In the opening session of the workshop, Tsokas recalled, Ameer had described the field of regenerative engineering as stemming from three
___________________
1 See https://www.fda.gov/downloads/biologicsbloodvaccines/guidancecomplianceregulatoryinformation/guidances/cellularandgenetherapy/ucm243392.pdf (accessed December 16, 2018).
revolutions in biomedicine: the molecular biology revolution, the genomics revolution, and convergence. Convergence brings together very disparate or separate disciplines that do not generally interact with each other, and the field of regenerative engineering is a field of convergence. As those in the workshop had learned, Tsokas said, there are many factors that can affect the variability of regenerative engineering products, including the immune system, disease state, and a range of other characteristics of the patient and the donor, the manufacturing process, and preservation techniques.
Three case studies had been discussed that illustrated some of the sources of variability and the challenges in developing these unique products, Tsokas said in review. Three successive panels had then considered patient sources of variability, variability in donor tissues and cells, and variability in the manufacturing process. Across these panels there was discussion of the role of the immune system in tissue regeneration, the unique supply chain for regenerative medicine products, and the importance of designing quality into the processes from start to finish. Another panel then discussed clinical trials and the regulatory approval pathway for regenerative engineering products, especially metrics and outcomes measures.
This workshop and the discussions throughout the day had fostered dialogue among the experts in this room from diverse sectors working in the field, Tsokas said. Participants had increased their understanding of the issues surrounding the sources of variability for regenerative medicine products and advanced discussions concerning how to account for and address these variables to provide a quality product for patient use. Key points that individual speakers made concerning these issues are provided in Box 7-1.
One of the issues that those in the workshop had debated, Tsokas said, was the variability in manufacturing processes and whether the process is the product or the product is the product. There are implications for each conclusion, she said,
but really it’s how you can control the variability. It’s one thing to say [that you are] starting with a variable patient population, but if you are starting with a variable product, you really do need to make sure that you are controlling the manufacturing process because that is what is going to ensure the quality of the product.
In the end, she said, the product quality is fundamental to making sure that good therapies are being provided that will address the unmet medical needs of patients.