Potency Assays in CAR-T and Cell Therapy BLAs: CBER Precedent and Review Deficiencies

Potency assays in CAR-T and cell therapy BLAs: what CBER accepted and where reviewers pushed back

Key takeaways

None of the licensed CAR-T products reviewed here relies on CAR expression alone for potency. Release specifications pair CAR expression and viability with an antigen-specific functional readout. Examples are IFN-γ release against CD19-expressing target cells for KYMRIAH ⁹¹⁹⁹²⁰, a "CD19-specific" redacted assay for AUCATZYL ⁷⁷³, and flow-cytometric cytotoxicity for the TCR-T product TECELRA ⁴²⁴. FDA's CAR T guidance treats CAR expression as an identity measure and asks for potency testing that measures biological activity ⁷⁹⁹⁸⁰⁰.
Most products use a potency matrix rather than a single assay. Examples include RYONCIL (two complementary assays plus viability and cell concentration) ⁶⁶⁰⁶⁶³, LANTIDRA (glucose stimulation index, islet yield, viability) ⁷⁵⁴⁷⁵⁶⁷⁵⁷, CASGEVY (on-target editing frequency plus two redacted assays) ⁶⁹⁷, and the lentiviral HSC products ⁶⁰⁵⁵⁶⁹⁶²³.
In CBER reviews, potency-related deficiencies fall into seven recurring groups:
- a single assay judged insufficient to control biological activity (CARVYKTI) ⁷¹⁹⁷²⁰
- a surrogate assay that lacked a demonstrated link to mechanism or clinical outcome (RYONCIL) ⁶⁶⁴⁶⁶⁷
- acceptance criteria not anchored to clinical lots (TECELRA, TREGZI, CASGEVY) ⁹⁸²⁵⁹⁶⁶⁹⁹
- validation and robustness gaps (KRESLADI, LYFGENIA) ⁶²⁶⁷⁰⁸
- stability-indicating potency questions (YESCARTA, CARVYKTI) ⁸⁵⁰⁷²³
- sampling and assay-execution problems (LANTIDRA) ⁷⁵⁹⁷⁶³
- release assays not yet linked to an orthogonal functional assay (TREGZI) ⁵⁹³
A late fix to the potency assay can push back the action date. CARVYKTI's replacement potency assay arrived late enough to trigger a major amendment, which moved the action due date from November 29, 2021 to February 28, 2022 ⁵⁰⁹⁹⁹⁴⁹⁹⁵.

The regulatory framework CBER applies

2011 final guidance, Potency Tests for Cellular and Gene Therapy Products.

Each lot must be tested for potency before release, and potency limits should help ensure lots are biologically active and consistently manufactured ⁸³²⁸³³.
The preferred approach is a quantitative bioassay that reflects relevant biological properties and, ideally, the mechanism of action ⁸³⁴⁸³⁵.
A matrix of complementary assays is acceptable when a single assay is insufficient. Qualitative assays used for release, stability, or comparability should be accompanied by at least one quantitative assay ⁸³⁷.
A non-biological surrogate assay needs data showing that it correlates with biological activity and can distinguish active from inactive or degraded product ⁸³⁸⁸³⁹.
The BLA must describe a validated potency assay or matrix. Acceptance criteria should reflect the potency of the lots used in pivotal studies ⁸⁴³.

2023 draft guidance, Potency Assurance for Cellular and Gene Therapy Products. This draft moves from a "release test" model to a potency assurance strategy. That strategy combines material controls, process-parameter controls, in-process testing, and lot-release assays ⁹⁶⁰.

For most CGT products, release testing should include at least one bioassay related to the intended therapeutic effect. The bioassay does not need to reproduce the full mechanism of action ⁹⁶⁹⁸²⁰⁹⁷⁰.
Acceptance criteria should have a quantitative lower limit that rejects sub-potent lots, and an upper limit where excess activity is a safety risk. "Report only" criteria are not recommended for release assays ⁸¹⁸.
Lot-release and in-process assays must be validated before a BLA ⁹⁶¹⁸³⁰.
Release criteria for a licensed product should keep commercial lots consistent with the lots that supplied the primary evidence of effectiveness ⁹⁷².

2024 final guidance, Considerations for the Development of CAR T Cell Products.

CAR expression is listed as an identity measure. Potency testing may use orthogonal methods such as cell killing and cytokine secretion ⁷⁹⁹⁸⁰⁰.
Bispecific CARs need activity assessed against each target antigen ⁸⁰¹.
Lot-release and stability assays should be qualified before effectiveness-supporting studies and validated to support a BLA ⁸⁰²⁸⁰³.

2026 draft guidance, Potency Assessment of Active Immunotherapy Products.

Quantitative bioassays are the most direct assessment, and complementary assays should be used when one test cannot cover all potency-related CQAs ⁹⁵¹⁹⁵³.
For live cell-based products, the draft asks for a measure of relevant cellular activity, with cell viability included in the overall assessment ⁹⁵⁴.

Potency assays in licensed CAR-T and TCR-T products

Product (BLA)	Potency-related release testing described in CBER review	Notable review observations
KYMRIAH (125646)	IFN-γ release in response to CD19-expressing target cells; CAR expression by flow cytometry; transduction efficiency by CAR-qPCR ⁹¹⁹⁹²⁰	IR on flow cytometry compensation, fluorescence-minus-one controls, and whether analyst adjustments changed release outcomes ⁹²⁴⁹²⁵; 2022 CMC commitment to revise the % viable T cell criterion ⁹²⁷
YESCARTA (125643)	Cell viability, anti-CD19 CAR expression, and a redacted potency test ⁸⁴⁸⁸⁴⁹	Large drop in a redacted potency result on stability compared with release; CBER said a drop of that size would exceed normal assay variation ⁸⁵⁰
TECARTUS (125703)	Cell viability, anti-CD19 CAR expression, and a third redacted potency test ⁹³³⁹³⁴	Methods and validations found adequate; no potency-specific deficiency in the reviewed pages ⁹³⁵
BREYANZI (125714)	Potency tested separately on the CD4 and CD8 components; release confirms CAR identity and CD19-specific activation ⁹³⁸⁹³⁹	Potency test and criteria redacted ³⁷⁶
ABECMA (125736)	Potency test on final formulated product (method redacted) ⁵⁵⁵⁵⁵⁶; CD137 activation listed among release attributes ⁴⁷³	Exploratory analysis found higher antigen-specific in vitro IFN-γ among attributes possibly associated with higher Cmax and AUC0-28d, with no strong overall relationship ⁴⁷⁴
CARVYKTI (125746)	CAR expression from viable T cells (potency/identity) plus a separate redacted potency assay that replaced the original commercial assay ⁷¹⁸⁵⁰⁹	Original approach judged insufficient (see deficiencies below) ⁷¹⁹⁷²⁰
AUCATZYL (125813)	One potency test, "CD19-specific (b)(4)"; CAR expression is listed separately for identity and quantity ⁷⁷³	Potency criterion set by tolerance interval because potency data were limited ⁷⁷⁴
TECELRA (125789)	Flow-cytometric cytotoxicity: MAGE-A4 TCR+ T cells killing T2 targets presenting MAGE-A4 peptide, compared against a reference ⁴⁰²⁴²⁴⁴²⁵	Lower limit revised twice in response to IRs ⁹⁸²⁹⁸³
AMTAGVI (125773, TIL)	Seven potency-related entries, six redacted, plus dose (total viable cells ≥7.5 × 10⁹) ⁶⁷⁸	Methods found adequate; formal meeting required before adding quality attributes or making major manufacturing changes ⁶⁸⁰⁶⁸¹

Nonclinical functional work in these files generally pairs co-culture cytotoxicity with cytokine release. For CARVYKTI, this meant BCMA-expressing targets at different effector-to-target ratios with IFN-γ and TNF-α readouts ⁴⁶⁴⁴⁶⁵. For AUCATZYL, it meant target-specific killing, cytokine secretion, and antigen-specific proliferation ⁴⁷⁰⁴⁷¹. The specific release assay was typically selected from this characterization package, which matches the guidance recommendation ⁸⁰⁰.

Potency assays in other licensed cell therapies

Product	Potency approach	Source
PROVENGE (sipuleucel-T)	CD54 upregulation ratio, a marker of APC activation; product at or above the median ratio (23.3) was associated with better survival in a simple analysis (HR 0.79)	⁹⁹⁹¹⁰⁰⁰
RYONCIL (remestemcel-L)	Final matrix of two complementary assays (one redacted, plus an optimized IL-2Rα inhibition bioassay), with viability and cell concentration (≥6.68 × 10⁶ cells/mL) under potency	⁶⁶⁰⁶⁶¹⁶⁶³
LANTIDRA (donislecel)	Glucose stimulation index (insulin ELISA, high/low glucose ratio ≥1), islet yield (≥5,000 EIN/kg initial), viability (≥70%)	⁷⁵⁴⁷⁵⁵⁷⁵⁶⁷⁵⁷
TREGZI (Orca Biosystems)	Redacted potency test plus viability, % viable Treg, and viable Treg dose; a traditional Treg suppression assay was judged impractical for release	⁵⁹²⁵⁹³⁵⁹⁴
ZEVASKYN (prademagene zamikeracel)	Four redacted potency entries, sampled during manufacturing because of the short shelf life	⁵⁶⁵⁵⁶⁶
ENCELTO (revakinagene taroretcel)	Redacted potency test; rhCNTF expression and viability also listed among release tests	⁶¹²⁶¹³
Cord blood (DUCORD, Bloodworks, MD Anderson)	TNC, viability, viable CD34+ cells (>1.25 × 10⁶/unit); CFU assay for MD Anderson	⁸⁸⁸⁸⁸⁹⁸⁹¹⁸⁹²
CASGEVY (exa-cel)	On-target editing frequency plus two redacted potency assays	⁶⁹⁷
ZYNTEGLO / LYFGENIA / SKYSONA	Each uses VCN and % LVV+ cells, plus a product-specific readout: transgene protein expression and CFC (ZYNTEGLO), transgene protein expression (LYFGENIA), %ALDP+ cells (SKYSONA)	⁶⁰⁵⁷⁰⁷⁶⁸⁶
LENMELDY / KRESLADI	Viability, VCN, transduction efficiency, ARSA activity (LENMELDY); neutrophil adhesion assay (KRESLADI)	⁵⁶⁹⁶²³

Potency-related deficiencies in CBER reviews

1. A single assay that does not control biological activity

CARVYKTI is the clearest CAR-T example.

In IR #11 (July 9, 2021), CBER said the then-current calculation of lot-release potency did not sufficiently control biological activity. It also said the proposed commercial assay was unacceptable as a standalone potency assay (IRs #11 and #20) ⁷¹⁹⁷²⁰.
FDA tentatively considered an alternative assay. Before deciding, it asked for the method description, validation, specifications, and batch results ⁷²⁰⁹⁹³.
Janssen replaced the original assay. Validation at the commercial testing site and updated batch analyses arrived in Amendment 56, which was designated a major amendment and extended the review clock ⁵⁰⁹⁹⁹⁴⁹⁹⁵.
CBER ultimately considered its potency concerns resolved. The new assay is validated at two sites, and any additional site needs validation data before it can test ⁵⁰⁹.

For CASGEVY, CBER had asked before Phase 3 for biologically relevant cellular assays in addition to editing-efficiency assays. At BLA submission, Vertex proposed a single potency measure. The reviewer instead accepted a combination of assays for the commercial specification ⁵⁴¹⁵⁴²⁶⁹⁸.

2. Surrogate assays without a demonstrated link to mechanism or outcome

RYONCIL's review history is the most detailed potency case in the CBER record.

The 2020 application relied on a non-biological assay of TNFR1 content ⁶⁵⁷. CBER found the TNFR1-survival analyses confounded by differences in populations and manufacturing processes. Knockdown data showed that inhibition of T cell proliferation was independent of TNFR1 ⁶⁶⁴⁶⁶⁵.
The complete response asked Mesoblast to:
- identify every assay it considered a potency test
- justify how those tests ensure each lot's activity
- show a statistically meaningful relationship to clinical outcomes, surrogate markers, or a relevant in vitro activity ⁶⁶⁷⁶⁶⁸
Reviewers also questioned assay variability. They noted that a manufacturing-related loss of potency had not been ruled out ⁶⁶⁹⁶⁷⁰.
The 2024 resubmission resolved these deficiencies with two complementary assays. The reviewer still noted that additional quantitative potency assays would be needed in the future ⁶⁷⁴.

The absence of clinical correlation is not always disqualifying. For CASGEVY, CBER noted that correlations with clinical outcomes were not performed, yet it accepted the potency specification ⁶⁹⁸.

3. Acceptance criteria not anchored to clinical lots

TECELRA. After IR #7, FDA noted that the revised lower potency limit relied on a lot from a patient who did not respond. After IR #8, the limit was reset to the lowest value associated with a reasonable expectation of benefit, adjusted for inter-assay variability ⁹⁸²⁹⁸³.
TREGZI. CBER objected to widening dose, purity, and potency criteria beyond clinically demonstrated ranges and set its own criteria in CMC IR #32. The applicant accepted them ⁵⁹⁶⁵⁹⁷.
CASGEVY. CMC IRs #8 to #10 tightened on-target editing frequency, viable cell count, and viability. The editing-frequency criterion was justified by the lowest value in a clinical lot with acceptable safety and efficacy ⁶⁹⁹⁷⁰⁰.
RYONCIL. CBER asked for both a minimum and a maximum TNFR1 limit consistent with pivotal lots, and for a higher minimum on the IL-2Rα assay ⁶⁷².
ZYNTEGLO. VCN, protein expression, and CFC limits were tied to values seen in patients who achieved transfusion independence ⁶⁰⁶.

4. Validation, robustness, and site-specific limits

KRESLADI. Rocket committed to prospectively revalidate the drug product potency assay, with a final report due March 31, 2027. CBER found the release methods adequate only together with the assay commitments ⁶²⁴⁶²⁶.
LYFGENIA. A redacted drug product assay had inadequate robustness validation, and robustness studies are due as a PMC ⁷⁰⁸.
RYONCIL. CBER asked for additional robustness data for potency assay validation ⁶⁷¹.
KYMRIAH. A flow cytometry IR covered compensation and gating controls for CAR+ percentages ⁹²⁴⁹²⁵. Another questioned an MOI calculation that assumed linearity despite validation data showing a nonlinear relationship ⁹²⁶.
DUCORD. Issues with the viable CD34+ assay were raised at mid-cycle and resolved through validation ⁸⁹⁰.

5. Stability-indicating potency

YESCARTA. CBER could not tell whether a large potency drop after freeze-thaw was a one-time effect or a trend. Kite committed to additional stability studies ⁸⁵⁰⁸⁵².
CARVYKTI. Replacing the potency assay left ongoing stability lots with no time-zero data. Janssen agreed to place three more lots on long-term stability ⁷²³⁹⁹⁶.
TREGZI. Potency fell below the proposed commercial specification at 72 hours. The reviewer accepted 72-hour storage based on stable viability and clinical experience ⁵⁹⁸⁵⁹⁹.
CASGEVY. A shipping-validation PMC was required, partly because potency-related attributes had not been evaluated ⁷⁰⁴.

6. Sampling representativeness and assay execution

LANTIDRA. Glucose stimulation index samples came only from the top, pre-culture islet fraction. As a result, CBER could not assess final product potency ⁷⁵⁹⁷⁶⁰. FDA said testing in the final container was still important even if results arrived after transplant ⁷⁶².
LANTIDRA, CRL. A complete response deficiency cited inadequate QC operator training on the full GSI procedure ⁷⁶³.
ZEVASKYN. Potency samples are taken during manufacturing because of testing limits and the short shelf life ⁵⁶⁶.

7. Orthogonal functional assays

For TREGZI, CBER asked the applicant to correlate the release assay with a traditional Treg function assay. That function assay was moved into process validation, stability, and comparability work, and it remained unqualified ⁵⁹³.

Implications for sponsors

Plan for a functional readout in the release specification. CAR expression, viability, and VCN are accepted as parts of a matrix. Reviews and guidance consistently expect a measure of biological activity tied to mechanism as well ⁸⁰⁰⁹⁶⁹⁷¹⁹.
Set potency limits from clinical lots, especially responders. IRs on TECELRA, TREGZI, and CASGEVY all pushed acceptance criteria back toward clinical experience ⁹⁸²⁵⁹⁶⁶⁹⁹.
Lock and validate the commercial potency assay before filing. CARVYKTI shows that replacing an assay during review can cost a major amendment and new stability commitments ⁹⁹⁴⁷²³.
Build stability data with the commercial potency assay from the start. Assay changes and freeze-thaw effects led to post-approval stability commitments for CARVYKTI and YESCARTA ⁷²³⁸⁵².
Expect redactions to hide most specifics. For most licensed CAR-T products, the method, target cell, and numerical acceptance criterion of the functional potency assay are redacted (b)(4) in the public review ⁸⁴⁸⁹³³⁷⁷³⁵⁵⁵. Precedent analysis therefore depends on review structure, IR history, and PMC language more than on assay parameters.