Industries / Frontier science / Cell and gene therapy
Making each dose
This work turns a patient's or a donor's cells into a dose that passes its release tests, and gets it back to the patient in time.
5
Milestones achieved, from 2016 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2016
Automated CAR-T in one machine
A University College London team ran the whole CAR-T process (cell selection, activation, adding the gene with a lentivirus, and growth) inside Miltenyi Biotec's CliniMACS Prodigy, a machine that keeps cells in a closed, single-use tubing set and feeds them automatically. Using blood from healthy volunteers, the cells matched hand-made batches for gene transfer, cell type and killing, and the yield was enough for expected doses.
Lessona closed machine lets a hospital without a large clean-room team make cells, which the authors said had been limited to a few centres.
Credit
Ulrike Mock, Martin Pule, Adrian Thrasher, Waseem Qasim and colleagues (University College London), with Miltenyi Biotec staff
2018
Doses outside the label
In July 2018, two months after Kymriah's US approval for diffuse large B-cell lymphoma, Novartis said some doses made for those patients were not meeting their commercial specifications, the release tests set at approval. The FDA had required at least 80% of the T cells in a dose to be alive, against the 70% Novartis aimed for in its trials. Novartis gave such doses free through an expanded-access programme. In December 2019 BioPharma Dive reported that about 10% of the time Novartis could not ship a dose at all, and that real-world data presented at the American Society of Hematology meeting found that 29 of 102 lymphoma patients and 14 of 92 leukaemia patients had received doses below 80%, with no sign that they fared worse.
Lessonrelease limits fixed at approval decide which doses can be sold, so they need data linking them to how patients do.
Credit
Novartis; FDA
2022
Too few manufacturing slots
After the first CAR-T for multiple myeloma, ide-cel, was approved in March 2021, the makers of the two myeloma CAR-T products gave each centre a small number of production slots a month. A Mayo Clinic doctor said the makers blamed a lack of commercial-grade viral vector, the disabled virus that carries the CAR gene. At the University of Kansas, 40 patients were eligible from March to July 2021; 24 got a slot and 16 did not. Median survival was 9 months for those who did not; it was not reached for those who did. A Moffitt Cancer Center doctor said in 2022 that its waiting list was at times over 100 patients.
Lessonmaking capacity and vector supply, not only the medicine, decided who was treated.
Credit
Nausheen Ahmed, Leyla Shune and colleagues (University of Kansas Cancer Center); The ASCO Post
2025
Platform status granted, then revoked
In June 2025 the FDA gave its first platform technology designation to Sarepta's AAVrh74 viral carrier. The designation lets a company reuse data from an approved product that shares the same well-understood, reproducible technology, to speed review and manufacturing checks for the next one. On 18 July 2025 the FDA revoked it after three deaths that appeared to result from acute liver failure in people given therapies built on that carrier, saying the evidence no longer showed it could be used in more than one drug without harming safety.
Lessona shared platform shares its risks as well as its data.
Credit
FDA; Sarepta Therapeutics
2025
Point-of-care manufacturing becomes law
On 23 July 2025 UK rules came into force that let medicines be made where the patient is treated, such as a hospital, under a new manufacturer's licence. One licensed control site supervises many places of manufacture, each working to a written master file. The UK government calls it a world first, and says a cancer patient's immune cells could be collected, changed and returned within days.
Credit
Medicines and Healthcare products Regulatory Agency (MHRA); UK Department of Health and Social Care
Watch for
CAR-T made in two days
Novartis says its T-Charge process makes a CD19 CAR-T, rapcabtagene autoleucel, in under two days, against the longer growth periods of earlier products, and that this keeps more naive and stem-like T cells, the kind that multiply and persist best. In its early lymphoma trial, doses 25 times lower than Kymriah gave response rates of 75% and 80%. An October 2025 conference abstract from a lupus study of 21 patients still described it as rapidly made. Novartis paused its autoimmune trials of the product in August 2026 after three deaths; two cancer trials continue.
Achieved whena regulator approves a CAR-T made in under two days, or an outside group reports the same making time and cell quality.
Claimed by
Novartis, September 2023, restated October 2025
Plant for donor iPS cells
Sumitomo Pharma says its SMaRT plant in Suita, Osaka, is the world's first commercial-scale plant for medicines made from donor iPS cells, and that its partner S-RACMO will make Amchepry for Parkinson's disease there, from cell stock supplied by the CiRA Foundation.
Achieved whenthe plant supplies doses for routine use and a regulator's inspection or an outside report confirms it.
Claimed by
Sumitomo Pharma and RACTHERA, March 2026
Open
01
A public vein-to-vein table
A free table of vein-to-vein time, the days from collecting a patient's cells to infusing the finished dose, for every approved CAR-T and T-cell product. Each row gives product, disease, country, year, the median reported, the exact definition the source used, and a link.
The briefCredit · unclaimed02
A table of failed doses
A free table, one row per published report, of how often making a CAR-T or other T-cell dose failed or gave a dose outside its release limits: product, disease, country, years, number of patients, number of failures, the reasons given (such as poor growth, low viability or low purity) and what happened next (remade, given anyway or not given). Every figure is a count taken from a paper, label or conference abstract, with its source.
The briefCredit · unclaimed
Open 01 · the brief
A public vein-to-vein table
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A free table of vein-to-vein time, the days from collecting a patient's cells to infusing the finished dose, for every approved CAR-T and T-cell product. Each row gives product, disease, country, year, the median reported, the exact definition the source used, and a link.
Why it is still open
A 2025 review and meta-analysis by Locke and colleagues compared three products in one lymphoma, large B-cell lymphoma (medians of 30.6 days for axi-cel, 35.9 for liso-cel and 48.4 for tisa-cel), and found that in patients given axi-cel, waits of 40 days or more went with fewer complete responses. A Kansas study reported 42 days for a myeloma product in 2021. A Europe PMC search of paper titles for "vein-to-vein" found six papers, none a table across all products and countries.
Built with what exists
The Europe PMC and ClinicalTrials.gov APIs; FDA and EMA product documents read with pdftotext; a large language model such as Claude to find candidate figures, with a person confirming each one; a CSV in a public repository.
First step
Enter the figures from the 2025 meta-analysis and the Kansas study into a CSV, with each source's definition of the start and end points.
Done looks like
Every FDA-approved CAR-T and T-cell product has at least one sourced vein-to-vein figure or is marked "not reported", and a second person's check of 30 random figures finds at least 95% match their sources.
Open 02 · the brief
A table of failed doses
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A free table, one row per published report, of how often making a CAR-T or other T-cell dose failed or gave a dose outside its release limits: product, disease, country, years, number of patients, number of failures, the reasons given (such as poor growth, low viability or low purity) and what happened next (remade, given anyway or not given). Every figure is a count taken from a paper, label or conference abstract, with its source.
Why it is still open
Reports come one country or one survey at a time. The UK National CAR T Panel found manufacturing failure in 38 of 981 patients (3.87%) approved for CAR-T for large B-cell lymphoma. The EBMT, the European society for blood and marrow transplantation, surveyed its centres on failures and out-of-specification doses in 2025. No table across products and countries turned up in searches.
Built with what exists
The Europe PMC API; FDA labels and review documents read with pdftotext; a large language model such as Claude to find candidate figures and quote the sentence each came from, with a person confirming each one; a CSV file in a public GitHub repository.
First step
Search Europe PMC titles and abstracts for "manufacturing failure" with "CAR T", and enter the first ten reports with each one's definition of failure.
Done looks like
Every FDA-approved CAR-T product has at least one sourced failure figure or is marked "not reported", the UK panel's 38 of 981 is reproduced from its paper, and a second person's check of 30 random figures finds at least 95% match their sources.