Release specifications set the minimum quality that every lot of a cell therapy has to meet before it can be administered. They are also among the most heavily negotiated parts of a CBER biologics license application. CMC and regulatory teams preparing a BLA need to know how FDA has handled viability, cell dose, identity and purity in earlier licensures. That precedent helps them set defensible acceptance criteria, anticipate information requests, and match their strength definitions to what reviewers have already accepted.
The analysis below draws on public CBER review memos and summary basis for regulatory action documents for licensed cell therapy products, including CAR T, TIL, TCR, gene-edited, islet and cord blood products. For each attribute, it covers how strength and dose are defined, which release tests are listed, where FDA has disclosed numeric criteria and where they are redacted, and how disclosure differs between genetically modified and non-genetically modified products.
Cell therapy release specifications in FDA BLA reviews: viability, cell dose, identity and purity
Most licensed cell therapies have a release specification table in the public CBER review record. The numeric acceptance criteria for viability, purity and potency are usually redacted as (b)(4) in those tables. FDA does disclose two other things fairly consistently: the definition of strength (dose), with numeric ranges, and the pass/fail safety criteria (sterility, mycoplasma). Unredacted numeric viability limits appear mainly for non-genetically modified products, such as allogeneic islets and licensed cord blood. The sections below set out what the review memos and summary basis documents show, attribute by attribute.
Key findings
- Dose is defined by the active cell population, not total cells. For CAR T products, strength is expressed as CAR-positive viable T cells. Kymriah calculates it as % CAR expression × viable-cell concentration × volume per dose ÷ 100 292. Yescarta and Tecartus use viable cell count combined with anti-CD19 CAR expression 25576.
- Viability limits are almost always redacted, in genetically modified and non-modified products alike. Kymriah, Yescarta, Tecartus, Abecma, Carvykti, Aucatzyl, Amtagvi, Tecelra, Casgevy and Ryoncil all list a viability test with a (b)(4) criterion 291254751622443335473124483.
- Where viability is disclosed, values sit at or above the 70% guidance floor. Lantidra requires ≥70% viable islets 60. The Bloodworks cord blood BLA requires >85% viable nucleated cells 541. FDA's 2008 somatic cell therapy CMC guidance describes 70% as generally the minimum acceptable viability specification, and a lower limit must be supported with data 613614.
- Identity usually pairs a transgene test with a cell-type test. Examples include CAR qPCR "positive" for Kymriah 296, CD34 expression plus on-target editing frequency for Casgevy 118119, and a CD166/CD105/CD45 panel for Ryoncil 485. This matches the CAR T guidance, which says HLA typing alone is insufficient 561.
- Safety criteria are the most transparent part of the table. Sterility "no growth" or "negative" and mycoplasma "negative" or "not detected" appear unredacted across products. Endotoxin limits are almost always redacted. Lantidra is the exception, with a disclosed ≤5 EU/kg/hour limit 68.
Strength (cell dose) specifications
FDA treats strength as a calculated release attribute. The calculation is tied to the potency-relevant cell fraction and, for weight-based products, to patient weight.
| Product | Dose/strength definition | Dose range or release criterion |
|---|---|---|
| Kymriah (tisagenlecleucel) | Transduced viable T cells | 0.2 to 5.0 × 10⁶/kg (≤50 kg); 0.1 to 2.5 × 10⁸ total (>50 kg) 292 |
| Yescarta (axicabtagene ciloleucel) | Viable CAR-positive T cells per kg | Release range (b)(4). Labeled target dose 2.0 × 10⁶/kg, maximum 2 × 10⁸ CAR T cells 255256257 |
| Tecartus (brexucabtagene autoleucel) | CAR-positive viable T cells from weight, viable concentration and CAR expression | Original BLA (MCL): 2 × 10⁶ CAR T cells/kg, capped at 2 × 10⁸ for patients ≥100 kg 768277. The later B-ALL indication uses 1 × 10⁶/kg, capped at 1 × 10⁸ |
| Breyanzi (lisocabtagene maraleucel) | CAR-positive viable T cells, CD8 and CD4 components dosed 1:1 | 50 to 110 × 10⁶ per dose in the original BLA review; release table strength criterion (b)(4) 442444445 |
| Abecma (idecabtagene vicleucel) | CAR-positive T cells from concentration, % CAR+ and bag volume | 300 to 460 × 10⁶ CAR-positive T cells 192021 |
| Carvykti (ciltacabtagene autoleucel) | CAR⁺ viable T cells, per kg (≤100 kg) or total (>100 kg) | 0.5 to 1.0 × 10⁶/kg; 0.5 to 1.0 × 10⁸ total 226219 |
| Aucatzyl | CD19 CAR-positive viable T cells | 410 × 10⁶ ± 25% 4445 |
| Tecelra (afamitresgene autoleucel) | MAGE-A4 TCR-positive T cells, calculated | ≥2,680.0 × 10⁶ cells; bags released are controlled to stay within the dose range 470471 |
| Amtagvi (lifileucel) | Total viable cells | 7.5 × 10⁹ to 72 × 10⁹ viable cells per dose 332333334 |
| Lantidra (donislecel) | Islet equivalents (EIN) per kg | ≥5,000 EIN/kg initial; ≥4,500 EIN/kg subsequent transplant 61 |
| Ryoncil (remestemcel-L) | Cell concentration | ≥6.68 × 10⁶ cells/mL 484 |
| Skysona | Total cell concentration | 4 to 30 × 10⁶ total cells/mL 462 |
| MACI | Cell density on collagen membrane | 500,000 to 1,000,000 cells/cm² 97 |
| Cord blood (Hemacord, Bloodworks) | Total nucleated cells and viable CD34+ per unit | TNC >5.0 × 10⁸/unit; viable CD34+ ≥1.25 × 10⁶/unit (Hemacord) 535537; >1.25 × 10⁶ viable CD34+ (Bloodworks) 541542 |
Two points matter when you benchmark against these values:
- Labeled dose and release criterion are not always the same thing. For Casgevy, the administered dose is at least 3 × 10⁶ viable CD34+ cells/kg, and one dose can be supplied from multiple lots. The release criterion for viable cell count itself is redacted 128129123. For Lyfgenia, the approval letter gives 1.7 to 20 × 10⁶ total cells/mL, including 1.4 to 20 × 10⁶ CD34+ cells/mL, and a minimum dose of 3.0 × 10⁶ CD34+ cells/kg. These values cannot be mapped to the redacted release-table rows 384380.
- Total cell count is often reported without a specification. Kymriah's total cell count is "report cells/mL" with no limit. The calculated viable-cell number does carry a (redacted) limit 293294295.
Viability
Every BLA reviewed includes a viability test. The disclosed criteria are limited:
- Lantidra: ≥70% viable islets by SYTO 13 Green/ethidium bromide stain with microscopic evaluation 60.
- Bloodworks HPC, Cord Blood: >85% viable nucleated cells 541. Hemacord lists CD45+ viability with a redacted criterion 536. MD Anderson's cord blood BLA redacts its nucleated-cell and CD34+ viability criteria 546.
- CAR T and gene-modified HSC products: Viability is listed, sometimes as a potency-related attribute (Yescarta, Tecartus), but the criterion is (b)(4) 25425975. For Breyanzi, FDA describes a finalized viability criterion but does not state the cutoff 440.
- Tissue-engineered products: Stratagraft has a viability assay with a redacted method and criterion 506507.
The guidance baseline is consistent. The 1998 somatic cell and gene therapy guidance asks for viability to be quantified with an established lower acceptable limit 612. Both the 2008 somatic cell CMC guidance and the 2020 gene therapy CMC guidance (for ex vivo modified cells) recommend at least 70% unless data support a lower limit 613615.
Identity
Identity tests in the approved tables fall into three patterns:
- Transgene detection: Kymriah uses CAR qPCR with a "positive" criterion 296. Yescarta and Tecartus detect scFv heavy-chain variable region, linker and CD28 sequences, with redacted criteria 25878. Carvykti is "positive for detection of CAR transgene" in the summary basis 227. Abecma requires anti-BCMA02 CAR-positive T cells to be detected 14. Aucatzyl requires CD19 CAR expression "detected" 46.
- Cell phenotype or edit: Casgevy requires "positive" for CD34 expression by flow cytometry and for on-target editing frequency by TIDE. The positivity thresholds are redacted 118119. Ryoncil tests CD166⁺, CD105⁺ and CD45⁺ cell content with redacted criteria 485; MSCs are conventionally CD166- and CD105-positive and CD45-negative. MACI requires "positive for chondrocyte cells" 92. For Zynteglo, βA-T87Q-globin protein expression serves as an identity test 411.
- Morphological or donor-matching identity: Lantidra combines packed tissue volume (≤10 cc) with dithizone-stained islet morphology 6263. Cord blood products rely on confirmatory HLA typing of the attached segment, ABO/Rh, and label matching 539543.
FDA's January 2024 CAR T guidance recommends testing at every development phase for both the transgene and the intended cellular composition, and states that HLA typing alone is insufficient 553561.
Purity and impurities
Purity attributes are listed but almost always redacted:
- Kymriah: % viable T cells and % viable CD19+ B cells (flow cytometry, CD45/CD19), plus residual CD3/CD28 beads by microscopy, all (b)(4) 297298299.
- Abecma: T-cell percentage, (b)(4) 15. Carvykti: phenotype measures including % NK cells, (b)(4) 229.
- Casgevy: CD34 purity by flow cytometry, (b)(4). The review treats nonviable and non-CD34+ cells as product-related impurities and Cas9 residuals as process-related impurities evaluated in characterization, not as listed release tests 120104.
- Ryoncil: Residual bovine serum albumin and residual trypsin, (b)(4) 487.
- Yescarta: Gentamicin as a process residual, (b)(4) 260.
- Amtagvi: The review states that residual viable melanoma cells are controlled by release testing. Neither the method nor the criterion is disclosed 346.
Vector copy number appears as a release test in the lentiviral HSC products, classified under potency/strength rather than purity. Zynteglo, Lyfgenia and Skysona each list VCN by qPCR with redacted criteria 408375462. The CAR T guidance asks sponsors to justify the VCN release criterion through a risk assessment 567. Several CAR T release tables (Yescarta, Tecartus, Abecma, Carvykti) do not list a separately named VCN specification 240703201.
Potency
Potency assays are product-specific, and their criteria are typically redacted:
- Kymriah: CAR expression by flow cytometry and IFN-γ release against CD19-expressing targets 301302303.
- Tecelra: Cytotoxic activity by a flow cytometry-based assay 475.
- Lantidra: One of the few fully disclosed potency criteria: glucose-stimulated insulin secretion ratio (high:low glucose) ≥1 by ELISA 64.
- Ryoncil: IL-2Rα inhibition, (b)(4) 486.
- Zynteglo: VCN, % LVV+ cells, colony-forming cells and βA-T87Q-globin expression, all (b)(4) 408409410.
- Casgevy: On-target editing frequency by TIDE plus two redacted assays 121122.
FDA's 2011 potency guidance does not prescribe a particular assay or acceptance criterion 616. The CAR T guidance suggests orthogonal measures such as target-cell killing, transduction efficiency and cytokine secretion 565.
Safety: sterility, mycoplasma, endotoxin and replication-competent virus
| Product | Sterility | Mycoplasma | Endotoxin | RCR/RCL |
|---|---|---|---|---|
| Kymriah | Negative 306 | Negative 306 | (b)(4) 306 | No detection of RCL 304305 |
| Yescarta | No growth 264 | Negative 262 | (b)(4) 261 | Negative (two tests) 262265266 |
| Tecartus | Negative 82 | Negative 81 | (b)(4) 83 | Not listed separately 70 |
| Breyanzi | No growth 450 | Not detected 451 | (b)(4) 452 | Not identified in release table 453454 |
| Abecma | No growth 24 | None detected 23 | (b)(4) 20 | Excluded from DP release testing 2526 |
| Carvykti | No growth 232 | Not detected 234 | (b)(4) 233 | (b)(4) 235 |
| Aucatzyl | No growth 50 | Negative 52 | (b)(4) 51 | (b)(4) 53 |
| Amtagvi | No growth 342 | Not detected 341 | (b)(4) 340 | Not applicable/not listed 307 |
| Casgevy | No growth 125 | Negative 126 | (b)(4) 127 | Not listed 104 |
| Lyfgenia | No growth 381 | None detected 383 | (b)(4) 382 | Not identified 349 |
| Lantidra | No growth in 14 days; fungal no growth in 28 days 6566 | Not stated 55 | ≤5 EU/kg/hour 68 | Not applicable |
| Ryoncil | Negative (in-process and final) 488 | Negative 488 | (b)(4) 489 | Not applicable |
The Abecma review is notable because RCL testing was excluded from drug product release instead of being assigned a criterion 2526. Kymriah sets its RCL qPCR limit from the assay LOQ 304305.
Products released before final sterility results
Several short-shelf-life products are released for infusion before the final compendial sterility result is available:
- Provenge: Release uses Gram stain, endotoxin content and a 2-day in-process sterility test. The 7-day sterility result is not available at infusion. Each dose contains at least 50 million autologous CD54⁺ cells activated with PAP-GM-CSF in 250 mL 199195196.
- MACI: Final sterility testing is done on conditioned medium, and the result is not available before administration 9419.
- Laviv: Requires a passing bulk drug substance sterility culture and a negative final-product Gram stain before shipping. Final sterility results can take up to 14 days 161.
- Gintuit: Requires passing in-process microbial tests before shipping, and the final sterility result can take up to 14 days 404.
Out-of-specification lots
In the Kymriah review, FDA describes lot release as the final quality check before commercial release 583584. FDA raised plans for disposing of occasional OOS lots in the commercial setting as an issue to discuss with Novartis 587. The available review text does not state the resulting commercial policy 575. One clinical batch that failed a color specification was released under an exception because its autologous erythrocytes were not considered a safety risk 588.
Practical implications for sponsors
- Define strength around the active population. Every approved CAR T and TCR-T product ties strength to the transgene-positive viable fraction 29225519470. Agree on the calculation (inputs, weight bands, caps) early, because it becomes a release specification.
- Expect to justify viability below 70%. The guidance floor is 70% 613615. The few disclosed BLA values are at or above it 60541.
- Build identity around two orthogonal elements, the transgene or edit plus the cell phenotype 561118119.
- Plan numerical criteria before pivotal studies. FDA's 2026 CMC flexibilities guidance accepts permissive early-phase criteria. It also states that release tests must have predefined acceptance criteria before Phase 2 or 3, and recommends discussing when to set numerical criteria for studies intended to provide substantial evidence of effectiveness 592.
- Benchmarking has limits. Because most numeric viability, purity and potency limits are redacted, public BLA records are most useful for test selection, dose definitions and safety criteria, not for copying numeric limits.