CAR-T Manufacturing — The Process Description That CBER Needs to Understand Your Drug

A CAR-T manufacturing process description is the only place in a CMC package where a reviewer can see whether you understand that every step modifies the drug. Most process descriptions treat it like a biologics upstream process.

The consequence of that misunderstanding is not a stylistic deficiency — it is a fundamental misrepresentation of how the drug is made. For CAR-T cell products under CBER OTP (Office of Therapeutic Products) jurisdiction, a Module 3 process description that reads like a monoclonal antibody upstream section will generate Information Request cycles, and in IND submissions progressing toward BLA, it will surface at pre-BLA meetings as a structural gap that requires retrospective process characterization data to close. The CMC package is not a laboratory notebook entry. It is the regulatory argument for why your manufacturing process reliably produces a safe and efficacious product — and for CAR-T cells, that argument must account for the fact that the product itself is transformed at each unit operation.

The Module 3 manufacturing process section for a CAR-T cell product must do something that no other drug class requires: describe a sequence of biological transformations — selection, activation, transduction, expansion, formulation — where each step changes the identity and quality of the product, and where the process parameters governing each step are simultaneously manufacturing controls and product quality determinants. That is the thesis this article will build toward, step by step, starting with the architecture of the process itself.

The CAR-T Manufacturing Process Architecture: From Apheresis to Infusion Product

The autologous CAR-T manufacturing process begins before a GMP facility ever receives a cell. It begins at the apheresis center, and the quality of the starting material — the leukapheresis product — will propagate forward through every subsequent unit operation to either constrain or enable the final drug product. FDA’s Chemistry, Manufacturing, and Controls (CMC) Information for Human Gene Therapy INDs (2020) guidance is explicit that the description of the manufacturing process must encompass the entire process from starting material through release testing, and for autologous products that chain begins with patient-derived cells collected under defined, controlled conditions. Leukapheresis yield targets — the number of viable CD3+ T cells collected per apheresis run — must be specified as a starting material acceptance criterion, not left as a site-variable outcome. A program that accepts any leukapheresis product without defined cell count, viability, and T cell composition thresholds has, at that moment, introduced a source of manufacturing variability that no downstream parameter can fully compensate.

Following apheresis, the process moves to T cell selection — typically CD4/CD8 positive selection using immunomagnetic separation — and then to cryopreservation of the starting material for centralized manufacturing. This step introduces a chain-of-identity obligation that is distinct from what is required for biologics: the process description must demonstrate, through traceability documentation and chain-of-identity records, that the cells used to manufacture a given lot can be traced unambiguously to the specific patient donor. This is not an administrative requirement. It is a product safety and identity control, and CBER has consistently flagged Module 3 sections that describe chain-of-identity in procedural terms rather than documenting the specific control points — labeling checkpoints, dual-verification steps, barcode reconciliation — that enforce it at the facility level.

The process flow diagram, which is a required component of the Module 3 process description, must not function as a schematic of equipment or rooms. It must display unit operations as sequential transformation steps, each accompanied by the in-process controls (IPCs) that govern the step and the critical process parameters (CPPs) that are being held within defined ranges. A diagram that shows “Transduction” as a single box with no associated parameter specifications or IPC checkpoints is a deficiency in its current form, regardless of how thoroughly the narrative describes the step. The diagram and the narrative must be co-referential.

T Cell Activation, Transduction, and Expansion: The Critical Steps With the Highest CMC Risk

T cell activation is the first step where the biological identity of the cell begins to change, and it is where many CAR-T CMC sections reveal that process development data were generated at research scale without translation to GMP-controlled parameter ranges. Activation using anti-CD3/anti-CD28 co-stimulation — whether via soluble antibodies, bead-conjugated systems, or membrane-bound constructs — depends critically on the ratio of activation reagent to T cells, the activation duration, and the cytokine environment. These are not background conditions. They are CPPs. If activation duration is described as “approximately 48–72 hours,” that is a research description. A GMP-controlled process description specifies the parameter as a defined range with justification for the limits and an IPC at the end of activation — typically a viability check and activation marker assessment — before proceeding to transduction.

Transduction efficiency defines the therapeutic index of the final product. A lentiviral or retroviral vector delivery step controlled by multiplicity of infection (MOI) is the mechanism by which the CAR transgene is introduced, and the MOI is the CPP that most directly determines the percentage of T cells expressing the CAR construct — the transduction efficiency — which is itself a critical quality attribute (CQA). MOI must be defined as a CPP with an operating range, and transduction efficiency must be specified as an IPC acceptance criterion with defined lower and upper limits. The lower limit is clinically driven: insufficient CAR expression produces a product that may not meet potency specifications. The upper limit is safety-driven: excessive MOI increases the risk of multiple integrations per cell, which carries insertional mutagenesis implications that CBER will expect to see addressed in the process description and in vector copy number data. If the MOI range and the transduction efficiency acceptance criterion do not appear in your Module 3 section as defined, justified parameters, that is a common deficiency pattern that FDA’s CMC Information for Human Gene Therapy INDs (2020) guidance flags directly through its expectation for characterization of the viral vector delivery step.

Expansion duration and conditions present a different risk category. T cell expansion over 10–14 days in the presence of recombinant cytokines — typically IL-2, or IL-7/IL-15 for specific CAR-T platforms — drives the fold expansion that produces sufficient cell numbers for the patient dose. But expansion duration also drives phenotypic differentiation. Prolonged expansion shifts T cell populations toward a terminally differentiated effector phenotype with reduced in vivo persistence potential. A Module 3 process description that defines expansion as a fixed duration without a harvesting decision rule tied to expansion fold-change and IPC data is not characterizing a manufacturing process — it is describing a protocol. The harvesting decision must be governed by defined expansion ratio limits, and those limits must be justified by process development data linking expansion parameters to the T cell phenotype CQAs — memory/effector ratio, CD4/CD8 ratio — that appear in the lot release specification.

The CPP-CQA Framework for CAR-T Manufacturing: What CBER Expects to See Characterized

ICH Q11, which governs the description of manufacturing processes for drug substances, establishes the expectation that process parameters be classified by their potential impact on product quality — that CPPs be identified, ranges be justified, and the linkage between process controls and critical quality attributes be explicit in the development and manufacturing sections of the CTD. For CAR-T cell products, the application of ICH Q11 is complicated by the fact that the drug substance and drug product are the same material, and the CQAs evolve across the manufacturing process rather than being established at a discrete synthesis step. FDA’s Considerations for the Development of Chimeric Antigen Receptor (CAR) T Cell Products guidance (finalized January 2024) acknowledges this complexity by framing the characterization expectation around the entire ex vivo manufacturing sequence, not merely the final formulated product.

The CPP-CQA framework that CBER expects to see characterized in Module 3 must link each unit operation to the CQAs it creates, maintains, or risks degrading. The transduction step creates CAR expression — transduction efficiency is the CQA, MOI is the CPP. The expansion step maintains viability and drives cell number — viability and fold expansion are the IPCs, cytokine concentration and culture duration are the CPPs. The formulation and cryopreservation step — which for most autologous CAR-T products uses DMSO at 5–10% as the cryoprotectant — determines the post-thaw viability and function of the infusion product. The DMSO concentration is not a trivial formulation variable: it must be specified with a defined range, and the post-thaw acceptance criteria for viability, CAR expression, and potency must be established against the same product that was cryopreserved under those conditions. A process description that specifies pre-freeze parameters carefully but leaves post-thaw acceptance criteria undefined has characterized only half the manufacturing step.

Process validation for CAR-T products under FDA’s Process Validation Guidance (2011) Stage 1 — Process Design — requires that the process be defined sufficiently to allow consistent manufacture at commercial scale. Applying Stage 1 to autologous cell therapy means that the process design phase must capture the full range of donor variability that will be encountered in the commercial population and demonstrate that the process can operate within defined parameters across that variability range. A clinical-stage program that has manufactured only from lymphocyte-rich, healthy donor leukapheresis products and has not characterized process performance from lymphopenic or heavily pretreated patient material has a Stage 1 gap that CBER will identify. The EMA/CAT Guidelines on GMP for ATMPs enforce a parallel expectation: the manufacturing process for an ATMP must be developed and validated with explicit consideration of the inherent variability of the starting material, and that characterization must appear in the process description and development report.

Writing the CAR-T Process Description That CBER Needs to Understand Your Drug

The XGene CAR-T Process Architecture and IPC Framework is a step-by-step manufacturing process description template for CAR-T CMC sections that defines each unit operation, its CPPs, its in-process controls with acceptance criteria, and the CQA linkage that justifies each control decision.

1. Unit Operation Mapping with Transformation Characterization: For each unit operation in the CAR-T manufacturing sequence — selection, activation, transduction, expansion, harvest, formulation, cryopreservation — document the specific biological transformation that occurs and the CQA it creates or affects, so that the process description makes the argument that controls were designed to manage that transformation, not merely to monitor the equipment.

2. CPP Definition with Operating Range Justification: For each identified CPP — including MOI for transduction, cytokine concentration and culture duration for expansion, and DMSO concentration for cryopreservation — define the operating range and provide the process development data rationale that justifies the range limits, linking upper and lower bounds to specific CQA impacts that would occur outside those boundaries.

3. IPC Checkpoint Specification with Acceptance Criteria: At each critical decision point in the manufacturing sequence — post-activation, post-transduction, at harvest, post-formulation, post-thaw — specify the in-process control test, the method, and the acceptance criterion as a defined numerical specification, not a qualitative descriptor, so that batch record review at pre-BLA inspection can verify that each manufacturing lot was governed by the same criteria used during process development.

4. Batch Record Co-Reference Alignment: Verify that every CPP range, IPC checkpoint, and acceptance criterion appearing in the Module 3 process description can be located by its exact parameter name and numerical value in the executed batch records for clinical manufacturing lots, establishing a traceable audit trail from the regulatory submission to the manufacturing floor record.

The output is a Module 3 process description section and supporting process flow diagram that function as an integrated regulatory argument — not a protocol summary — demonstrating to CBER reviewers that each manufacturing step is controlled, that each control decision is scientifically justified, and that the executed clinical manufacturing records confirm the process operated as described.

A CAR-T CMC package built on a weak process description creates a compounding liability. Deficiencies identified in Module 3 during IND review propagate into Phase 2 and Phase 3 manufacturing changes that require comparability data to bridge — data that would not have been necessary if the process had been characterized and documented at the required level from the start. By the time a program reaches pre-BLA readiness assessment, an underdeveloped process description is not correctable by revision alone: it requires retroactive process characterization studies, manufacturing history gap analysis, and in some cases prospective process validation batches that carry full GMP cost. The regulatory cost of a process description written at research-protocol depth is not a deficiency letter — it is a development timeline measured in additional years.

For your CAR-T manufacturing process, can you identify today the specific document that defines the acceptance criterion for transduction efficiency as an in-process control, the MOI range used during transduction, and the expansion ratio limits that define the harvesting decision — and confirm these same parameters appear in the executed batch records for your clinical lots?