N-of-1 and n-of-few: the custom-therapy route, and what it actually costs
also called n-of-1 trial · bespoke therapy · milasen · custom gene therapy
Two children have now received drugs designed for them alone: one an antisense drug made in about a year, one a gene-editing therapy made in about six months. That is the literal version of the PI's 'custom therapies' idea, and it is real, but each one required a whole academic centre working on one patient.
The same thing, for a scientist
The n-of-1 route is no longer hypothetical. Milasen, a splice-modulating antisense oligonucleotide designed against a single patient's private SVA insertion in MFSD8, went from genetic diagnosis to dosing in approximately one year under an individual IND. In 2025 a patient-specific in vivo base-editing therapy was designed, manufactured and administered to an infant with a severe urea-cycle disorder on a roughly six-month timeline. Both worked in the narrow sense. Neither produced generalisable efficacy evidence, both consumed the resources of a major academic centre for one child, and the model does not obviously transfer to a disease where the driver lesion is an oncogenic fusion rather than a loss-of-function allele.
The two real cases3 deeper
Milasen (2019). A child with neuronal ceroid lipofuscinosis 7 carried a private SVA retrotransposon insertion in MFSD8 causing aberrant splicing. A splice-modulating antisense oligonucleotide was designed specifically against that insertion, taken through toxicology, and administered under an individual investigational new drug application, with the interval from diagnosis to first dose approximately one year 1.
Patient-specific in vivo base editing (2025). An infant with a severe urea-cycle disorder received a bespoke in vivo base-editing therapy designed against the patient's own variant, manufactured and dosed on a timeline of months rather than years 2.
These are the existence proofs for 'custom therapies based off their custom biology'. They are also two patients.
Why the model does not transfer cleanly to KMT2A leukemia
Four differences, each of which matters.
- Direction of the lesion. Milasen and the base-editing case corrected or bypassed a loss of function in a gene the patient needed. A KMT2A fusion is a gain of function: an oncoprotein that must be removed or blocked, not restored. Antisense and base editing are far better suited to the former.
- Tempo. A one-year or six-month design-to-dose timeline is acceptable in a slowly progressive neurodegenerative or metabolic disease. Infant KMT2A-rearranged leukemia relapses mostly within a year of diagnosis, two thirds during treatment. A six-month bespoke manufacturing timeline is longer than the window in which most of these children relapse.
- Target heterogeneity within the patient. A germline variant is in every cell and does not change. A leukemic clone evolves under treatment, so a therapy designed against today's clone may miss tomorrow's, which is the standard subclonal-escape problem.
- Evidence value. An n-of-1 produces no generalisable efficacy evidence by construction. In a disease with ~48 eligible infants a year internationally, spending one of them on a study that cannot inform the next child is a real opportunity cost.
Where n-of-few genuinely fits this disease
Three places, all narrower than the headline idea and all defensible.
- A private breakpoint as a diagnostic and MRD target, not a drug target. Every KMT2A fusion has a patient-unique genomic junction, already used for MRD monitoring at sequence level 3. That is an n-of-1 assay, cheap and already routine, and it is the infrastructure a bespoke therapeutic strategy would need anyway.
- The fusion junction as a neoepitope. The graph already carries
target-fusion-junction-epitope. A junction-specific TCR or CAR is inherently patient-or-breakpoint-specific and would be an n-of-few product rather than an n-of-1 one, since breakpoints cluster. - Pooled n-of-1 designs. A series of n-of-1 experiments with a common protocol and endpoint can be meta-analysed, recovering some generalisable inference from individually bespoke interventions. This is the only version of the custom-therapy idea that also produces evidence, and it should be how the lab frames it.
The cost, stated honestly
No published per-patient cost exists for either milasen or the base-editing case; the teams have described the effort qualitatively rather than costing it. This node does not estimate it, because an invented figure would be worse than none.
What can be said from the record is that each required a major academic centre, a bespoke toxicology package, an individual IND and regulatory engagement, for one patient. Against a disease with ~160 US infant cases a year, a strategy requiring that per patient is not a platform; it is a series of heroic exceptions. The platform argument only becomes real if the shared component - one interface, one drug, many genotypes - carries most of the load, and the bespoke component is confined to diagnostics, monitoring and possibly the immunological target.
| design class | n-of-1 and n-of-few |
|---|
precedents
[
{
"case": "milasen antisense oligonucleotide",
"year": 2019,
"timeline": "~1 year diagnosis to dose",
"pmid": "31597037",
"title": "Patient-Customized Oligonucleotide Therapy for a Rare Genetic Disease.",
"authors": "Kim J, Hu C, Moufawad El Achkar C, et al.",
"venue": "N Engl J Med",
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"case": "patient-specific in vivo base editing",
"year": 2025,
"timeline": "months",
"pmid": "40373211",
"title": "Patient-Specific In Vivo Gene Editing to Treat a Rare Genetic Disease.",
"authors": "Musunuru K, Grandinette SA, Wang X, et al.",
"venue": "N Engl J Med",
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}
]transferability barriers
[ "gain-of-function rather than loss-of-function lesion", "relapse tempo faster than bespoke manufacturing", "subclonal evolution of the target", "no generalisable efficacy evidence" ]
defensible applications in kmt2a
[ "patient-specific breakpoint MRD assay", "fusion-junction neoepitope as an n-of-few target", "pooled n-of-1 designs with a common protocol" ]
not found
[ "published per-patient cost of any n-of-1 antisense or gene-editing therapy" ]
- Bespoke therapy consumes exactly the patients a generalisable trial needs. In a disease with 48 eligible infants a year internationally, every child treated on a one-off protocol is a child not contributing to an answer for the next one, and that trade is rarely made explicit to families.
- A therapy designed against a patient's specific fusion junction selects for subclones that have lost or altered that junction. The more precisely custom the therapy, the narrower the escape route it has to close.
- Framing the project as 'custom therapies' invites funders to expect n-of-1 deliverables, which are unaffordable at scale, when the fundable version of the idea is a shared interface serving many patients. The language matters for what gets funded.
- What is the actual per-patient cost and person-time of an n-of-1 antisense or base-editing programme? Nobody has published it, and the platform argument cannot be costed without it.
- Is there any bespoke modality fast enough to matter in a disease where two thirds of relapses occur on treatment, or does tempo rule the whole n-of-1 route out of frontline infant leukemia?
- Could a pooled n-of-1 design across KMT2A breakpoint-specific immunotherapies produce generalisable inference, and what would the common endpoint be?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
A splice-modulating antisense oligonucleotide (milasen) was designed against a single patient's private SVA insertion in MFSD8 causing neuronal ceroid lipofuscinosis 7, taken through toxicology and dosed under an individual IND approximately one year after the genetic diagnosis.
Patient-Customized Oligonucleotide Therapy for a Rare Genetic Disease · Kim J, Hu C, Moufawad El Achkar C, et al. · New England Journal of Medicine 2019What this evidence cannot show
A single patient. Clinical benefit was assessed against the patient's own prior trajectory, which is the only possible comparator and also the weakest one. No efficacy inference generalises.
Evidence that argues against this (1)
- The published assessment of this route says the field lacks the machinery to do it at scale. A genetic, regulatory and ethical review of n-of-1 antisense oligonucleotide development in Europe describes the area as 'academia-driven' and 'in need of systematic guidance and standards', notes that the current focus is on the 'more common' rare diseases leaving most patients without prospect of a disease-modifying treatment, and recommends that every single trial of therapy be registered in an ASO treatment registry and include generic outcomes so that results can be aggregated across n-of-1 cases. In other words, the milasen model as practised generates no cumulable evidence, and the authors treat that as the central defect rather than an acceptable cost. Abstract read via PubMed; the paper is open access (PMC9058873) and the full text was not read. Synofzik M et al. 2021
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A patient-specific in vivo gene-editing therapy was designed, manufactured and administered to treat an individual patient with a rare genetic disease, demonstrating that bespoke base editing can be delivered on a months-long timeline.
Patient-Specific In Vivo Gene Editing to Treat a Rare Genetic Disease · Musunuru K, et al. · New England Journal of Medicine 2025What this evidence cannot show
A single patient with short follow-up. Durability, off-target editing over time and the reproducibility of the manufacturing timeline outside one highly resourced centre are all unestablished. The author list and full clinical detail were not independently re-verified for this node beyond the title, journal, date and DOI.
Evidence that argues against this (1)
- One patient, one metabolic genetic disease, and a months-long timeline achieved by an exceptional team with regulatory attention that does not scale. It demonstrates feasibility for a correctable germline lesion in a non-malignant condition; a fusion oncoprotein in a proliferating malignant clone presents a different problem, since editing has to reach every leukaemic cell rather than a fraction of hepatocytes. Leukemia 2021
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Patient-specific KMT2A genomic breakpoint sequences derived from a diagnostic reference laboratory series of 3,401 acute leukemia patients are used directly to monitor minimal residual disease, translating a private lesion into a clinical assay.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
A referral-based diagnostic series. The MRD application is described by the authors as a translational use of their breakpoint data rather than validated in a prospective outcome study within this paper.
Evidence that argues against this (1)
- Searched and empty, recorded as a finding. On 2026-09-18 I queried the ClinicalTrials.gov API v2 for interventional studies combining n-of-1, individualized or patient-specific terms with KMT2A, MLL-rearranged or infant leukaemia and reviewed the top 25 records. None was an n-of-1 or patient-specific therapeutic study in KMT2A-rearranged leukaemia. The nearest registered analogue is NCT02670525, 'Matched Targeted Therapy For High-Risk Leukemias and Myelodysplastic Syndromes' (maximum age 30 years, status ACTIVE_NOT_RECRUITING as read on that date), which is a matched-therapy umbrella and not a bespoke per-patient product. So the n-of-1 route described in this entry has, as of 2026-09-18, no registered instance in this disease: patient-specific KMT2A breakpoints are used as an assay, not as a drug target. A searched-and-empty field and a blank one are different facts, and this is the former. US National Library of Medicine (registry query et al. 2026
Counter-evidence (Synofzik 2021, PMID 34591693; ClinicalTrials.gov query 2026-09-18 returning no match): the n-of-1 ASO field is described by its own practitioners as academia-driven and lacking systematic standards, producing no cumulable evidence; and as of 2026-09-18 there is no registered n-of-1 or patient-specific therapeutic study in KMT2A-rearranged leukaemia. Searched and empty, not blank.
the defensible n-of-few application in this disease
patient-unique breakpoint as an n-of-1 assay rather than an n-of-1 drug target
opportunity cost of bespoke treatment in a 48-patient-a-year population