There is a management instinct in drug development to organize work into functional lanes. Scientists design the product. Engineers build the process. Regulatory affairs translates both into submissions. Each group optimizes within its lane, and the program advances through defined gates.
That instinct is understandable. It works reasonably well for conventional biologics and extremely well for small molecules. In cell and gene therapy, it is one of the more reliable predictors of late-stage problems.
The reason is structural. In CGT, product design and process design are not parallel workstreams that converge at manufacturing. They are a single feedback loop, and every decision made on one side changes the constraints on the other. When that loop is managed as though it were two separate tracks, the disconnects compound quietly — until they surface as a batch failure, an unexpected CQA deviation, a regulatory question that the data cannot answer, or a Phase 3 readout that cannot be explained by the product profile in the IND.
Understanding why this happens, and what to do about it, requires looking at what product design and process design actually encompass in the context of CGT — and how the interaction between them differs from what developers experienced in earlier therapeutic modalities.
What Product Design Actually Covers in CGT
In a conventional biologic program, product design is primarily a scientific activity: define the target, select the molecule, characterize its binding profile, and iterate on potency and specificity. The product is the molecule, and the molecule is relatively stable — it does not change based on how it is grown.
In CGT, the product is not stable in that sense. A CAR-T cell product has a design, but it also has a phenotype, a persistence profile, an exhaustion trajectory, and a safety envelope — and all of those are influenced by how the cells were manufactured. An AAV gene therapy has a target serotype and a transgene cassette, but it also has a capsid full/empty ratio, a VCN, a residual host cell protein burden, and an immunogenicity profile — and all of those are influenced by the production platform and the purification process.
Product design in CGT encompasses hypothesis, mechanism of action, target, effective attributes, safety profile, and persistence. But it also encompasses how those attributes present in the final product — and that is where the boundary with process design begins to dissolve. The editing efficiency of a gene-engineered cell is a process outcome. The stability of transgene expression is a process outcome. Cell fitness for therapeutic benefit — whether the cells that reach the patient are capable of the sustained activity the clinical hypothesis requires — is a process outcome.
This means that every product design decision carries a process implication, and every process decision carries a product consequence. The two cannot be optimized independently.
What Process Design Actually Covers in CGT
On the process side, the picture is equally layered. Starting cell characteristics are a process input — but they are also a product determinant. For autologous therapies, the starting material is the patient’s own cells, collected under conditions the manufacturer cannot fully standardize. The health, immune status, and prior treatment history of that patient directly affect the cells that enter the manufacturing process, which directly affect the cells that emerge as the drug product. A process that was validated on healthy donor leukapheresis may not perform consistently on heavily pre-treated oncology patients.
For allogeneic therapies, starting material variability shifts to donor selection and screening — a different set of constraints, but equally consequential for what the process produces. The acceptance criteria for allogeneic donor starting materials must balance scientific rigor with supply chain reality, and those criteria are not merely regulatory checkbox items. They are load-bearing inputs to the entire downstream quality profile.
Beyond starting materials, process design covers cell expansion conditions, transduction or transfection procedures, hold times, purification steps, formulation, and stability testing. Each of these parameters can influence the final product’s safety, identity, purity, potency, and stability — the five CQA categories that will define the regulatory submission. Process changes at any point in the workflow require evaluation against the full CQA profile, because a change that looks localized to a single unit operation may propagate unexpected quality effects through the rest of the process.
This is precisely what the FDA means when it identifies changes in manufacturing as potential sources of change across all six product quality dimensions: safety, identity, quality, purity, potency, and stability. The list is not bureaucratic conservatism. It reflects the actual biological behavior of CGT products, which are more sensitive to process variation than any therapeutic category that preceded them.
Where the Feedback Loop Breaks Down in Practice
The most common manifestation of disconnected product and process management in CGT is the late-stage potency problem. Potency — the specific ability of the product to effect a given clinical result — is the most complex CQA in CGT because it is product-specific, mechanism-dependent, and analytically demanding. A valid, qualified potency assay for a CAR-T product requires understanding the mechanism of action well enough to measure it, developing a cell-based or functional assay that captures it reliably, and qualifying that assay to the standards required for lot release.
That work cannot begin until the product design is stable enough to define what is being measured, and the process is stable enough to produce a consistent product to measure it against. In programs where product design and process development are running on separate timelines, potency assay development routinely falls into the gap between them — started too late to support Phase 2 or Phase 3, or developed against a product profile that changes when a process optimization is made.
The same dynamic applies to characterization. The molecular, biochemical, immunologic, and phenotypic characterization of a CGT product at BLA requires a level of detail that cannot be generated against a moving target. If process changes during Phase 2 or Phase 3 alter the product’s phenotype or impurity profile, previously generated characterization data may not support the final submission. The sponsor has to decide whether to generate new data, argue comparability, or accept the regulatory uncertainty that comes with incomplete characterization.
Comparability itself — demonstrating that a post-change product is equivalent to the pre-change product — is a major operational burden in CGT precisely because the products are so biologically complex. A batch manufactured under the pre-change process and a batch manufactured under the post-change process may look similar on the CQA panel and behave differently in the clinic, or look different on the CQA panel and behave similarly. Establishing which attributes matter, in what ranges, and what comparability actually means for a given product requires the kind of deep integration between product knowledge and process understanding that only comes from managing them as a single system.
Table 7. Product design and process design attributes in CGT, organized by operational dimension. The gray footer row summarizes the feedback loop implication that applies across all dimensions: the two workstreams are biologically and analytically interdependent and cannot be managed as parallel-but-separate programs without generating late-stage quality and regulatory failures.
| Dimension | Product Design | Process Design |
|---|---|---|
| ↔︎ The Feedback Loop: How Each Side Affects the Other | ||
| Core question being answered | What does the therapy do and how does it do it? | How is the therapy reliably manufactured? |
| Key attributes governed | Hypothesis; MOA; target cell/antigen; effective attributes; safety profile; persistence characteristics | Starting cell characteristics; stable gene expression; gene editing efficiency; cell fitness for therapeutic benefit; purity/impurity control; reproducible quality; dose |
| Where variability originates | Mechanistic uncertainty; target biology; patient-specific immune environment | Donor-to-donor starting material variability; in-process parameters; scale-up; raw material lots |
| What changes to one side do to the other | Changing MOA target or CAR construct design changes the transduction and expansion parameters required to produce a functional product | Changing transduction protocol, expansion conditions, or purification steps can alter cell phenotype, VCN, potency profile — all product attributes |
| Potency assay dependency | Potency must measure the downstream biological effect of the product — MOA confirmation, tumor cell killing, transgene expression | Potency assay must be stable enough to measure consistently across process changes — requires the process to be producing a product with measurable, consistent functional activity |
| Regulatory touchpoint | FDA/EMA require MOA characterization, CQA definition, and downstream biological activity as part of potency (BLA/MAA) | FDA/EMA require CPP identification, in-process controls, and demonstration that process changes do not affect product quality (comparability data) |
| Most common failure mode | Late-stage discovery that the potency assay cannot measure the claimed MOA, because MOA was not sufficiently defined during product development | Late-stage process change that shifts cell phenotype or VCN in ways that existing analytical methods cannot detect — and that only becomes apparent in clinical outcomes |
| Feedback Loop Implication: Every row above has an interaction effect. Changes to either Product Design or Process Design must be evaluated against the full CQA profile — because the two are not independent workstreams. Managing them as separate organizational functions consistently produces late-stage quality failures that could have been anticipated at program inception. | ||
What Integrated Management Actually Looks Like
Managing product design and process design as a feedback loop requires organizational choices that do not come naturally to teams trained in conventional drug development.
It requires that process development scientists be in the room when mechanism of action hypotheses are being formulated — not to constrain the science, but to flag early which hypotheses will create manufacturing constraints that the program will eventually have to solve. A product design that requires highly specific cell subsets, or a particular activation state, or a manufacturing step with no validated scale-up path, is not just a scientific choice. It is a resource and timeline commitment that the development team should make with full visibility.
It requires that product scientists be in the room when process changes are being evaluated — not to slow down manufacturing optimization, but to assess what each change means for the product profile. A shift in transduction protocol that increases yield by 20 percent is a good outcome if the cells produced are equivalent in fitness and potency to the pre-change batch. It is a problem if the phenotype has shifted in ways that the existing analytical panel cannot detect.
It requires a shared analytical framework — a CQA map that both teams use as the connecting language between product knowledge and process performance. Every process parameter that can affect a CQA should be identified, documented, and monitored. Every product attribute that varies across batches should be traceable to a process variable. The control strategy that emerges from this framework is not merely a regulatory deliverable; it is the operational backbone of a manufacturing program that can produce a consistent, characterizable product at scale.
And it requires a regulatory strategy that reflects this integration. The FDA expects CGT sponsors to demonstrate that they understand the relationship between their manufacturing process and their product quality. Submissions that present product data and process data as parallel but disconnected sections — rather than as an integrated story about how the process produces the product — invite exactly the kinds of regulatory questions that slow programs down.
The programs that navigate CGT development successfully are those that internalize this integration early enough to build it into the organizational structure, the analytical roadmap, and the regulatory strategy before the late-stage consequences of managing them in isolation become the problem they are solving.
MKA Insights works with CGT developers at the intersection of analytical science, manufacturing strategy, and regulatory execution. If your program is at a stage where the connection between product design and process development needs to be built — or rebuilt — we bring the cross-functional perspective that this work requires.