The N-of-1 Trial as Infrastructure

The first individualized antisense oligonucleotide for a given gene target requires building everything from scratch. The second requires adapting it. The tenth requires adjusting it. By the hundredth, the infrastructure is mature, the safety database is deep, and the regulatory pathway is routine. Each person treated generates data that makes the next treatment faster, cheaper, and more certain.

This compounding effect is the central economic and scientific argument for individualized therapy. A drug made for one person is expensive in isolation. A drug made for one person whose data reduces the cost and risk of the next drug, and the next, and the next, is the foundation of a learning system that gets better with every use.

How the Acceleration Works

The FDA's Plausible Mechanism Framework explicitly contemplates this compounding logic. Master protocols allow multiple individualized therapies targeting different mutations in the same gene to be evaluated through a single regulatory application. The framework recognizes that if the mechanism of action is well characterized (exon skipping, splice correction, translation readthrough) and the chemistry class has an established safety profile, each new therapy within that framework carries less regulatory uncertainty than the previous one.

The practical acceleration unfolds in stages.

The first person treated with an ASO targeting a mutation in a given gene bears the full development burden. The mutation must be identified and characterized. The ASO must be designed, synthesized, and tested in cell models. Toxicology studies must establish a safety baseline from zero. GMP manufacturing must produce a clinical-grade batch. The regulatory submission must establish the mechanism, present the preclinical data, and justify first-in-human dosing. Timeline: 10 to 12 months. Cost: $1 to $3 million.

The second person, with a different mutation in the same gene, inherits data from the first. The chemistry class is the same. The delivery route is the same. The toxicology profile of the backbone modification is established. The regulatory pathway is precedented. The ASO design changes (different nucleotide sequence targeting a different mutation), but the manufacturing process, the toxicology framework, and the regulatory template carry over. Timeline: shorter. Cost: lower.

The fifth person, with yet another mutation in the same gene, benefits from a growing safety database across multiple treated individuals. The FDA's confidence in the mechanism increases with each treatment. The preclinical package can be further abbreviated. The regulatory review is faster because the reviewers have evaluated the same platform multiple times.

The twentieth person, possibly targeting a mutation in a different gene but using the same chemistry class and delivery route, inherits the entire platform's safety and manufacturing history. The mechanism is mature. The manufacturing is routine. The regulatory pathway is well established. The development cost per person approaches the manufacturing and clinical monitoring costs alone, with the design, toxicology, and regulatory overhead amortized across the preceding nineteen.

The Data That Compounds

The compounding does not happen automatically. It requires that the data from each treated person is captured, structured, and available to inform subsequent treatments.

If the first person's outcome data sits in one academic center's files, inaccessible to the clinician developing a therapy for the second person at a different institution, the learning stops. The second program repeats the toxicology. The third program repeats the regulatory preparation. Each program operates as if it were the first, because the data from prior programs is locked in institutional silos.

The data that matters falls into three categories.

Safety data: adverse events, laboratory abnormalities, immune responses, injection site reactions, organ toxicity markers. Each treated person's safety data expands the safety database for the chemistry class. A safety signal detected in one person can prevent harm in subsequent treatments. The absence of a safety signal across multiple treatments increases confidence for regulatory approval of the next.

Efficacy data: biomarker changes, protein production, clinical improvement, functional milestones. Each person's efficacy data contributes to the evidence base for the mechanism. If splice correction consistently restores protein production across multiple mutations in the same gene, the mechanistic validation strengthens. If one mutation responds differently, the data reveals which structural features of the target RNA predict response.

Pharmacokinetic and pharmacodynamic data: drug distribution, tissue penetration, dose-response relationships, duration of effect. Each person's PK/PD data refines the dosing framework for subsequent treatments. The optimal dose of an intrathecal ASO, the frequency of readministration, and the relationship between CSF drug levels and clinical response are established through cumulative data across treated individuals.

The Cross-Gene Extension

The compounding logic extends beyond a single gene. If ASOs targeting different mutations within Gene A all use the same chemistry class (2'-O-methoxyethyl modification, phosphorothioate backbone) and the same delivery route (intrathecal injection), the safety and manufacturing data are relevant to ASOs targeting mutations in Gene B using the same platform.

The intrathecal ASO safety database established through nusinersen (targeting SMN2 for SMA) is cited in regulatory submissions for ASO programs targeting entirely different genes for different neurological conditions. The chemistry is the same. The delivery is the same. The safety profile transfers.

This cross-gene transfer of platform data means that the value of each n-of-1 treatment extends beyond the specific gene and the specific condition. A child treated with an individualized ASO for CLN7 Batten disease (milasen) generates platform data relevant to individualized ASOs for other forms of Batten disease, for other neurodegenerative conditions, and for any condition where intrathecal ASO delivery is a viable approach.

The data does not compound within a disease. It compounds across the platform.

What Breaks the Flywheel

Three structural failures can prevent the compounding from occurring.

Data fragmentation: individualized therapy programs are currently conducted at a small number of academic centers (Boston Children's Hospital, University of Rochester, University of Pennsylvania, and others). Each center holds its own data. No shared repository aggregates safety, efficacy, and PK/PD data across all individualized ASO programs regardless of institution. A clinician at one center developing a therapy for a new mutation in a gene already targeted at another center may not have access to the prior program's data.

Sponsor discontinuation: if an academic center loses funding for its individualized therapy program, the follow-up data for previously treated individuals may stop being collected. The long-term safety data that would have strengthened the platform's evidence base is lost. The person treated under the program continues to receive the drug (or not), but the systematic data collection that makes their experience useful to others ceases.

Regulatory silos: if each individualized therapy is evaluated as a standalone IND rather than under a master protocol, the regulatory precedent does not accumulate efficiently. The FDA reviewer evaluating the fifth ASO targeting the fifth mutation in the same gene should be able to reference the prior four. If each submission is independent, with no formal mechanism linking them, the reviewer may require a full preclinical package for each.

The PMF framework addresses the regulatory silo problem. The data infrastructure problem, making each program's data available to inform subsequent programs, is not yet solved.

The Infrastructure Requirement

A data trust that holds outcome data from every individualized therapy program, structured for cross-program analysis, governed by the treated individuals and their families, and accessible to any clinician developing the next therapy, is the bearing that makes the flywheel turn.

The trust would hold safety data from every intrathecal ASO administered under a master protocol, regardless of the target gene or the treating institution. It would hold efficacy data stratified by mutation type, gene target, and mechanism of action. It would hold long-term follow-up data from every treated individual, for as long as that individual chooses to contribute.

The value of this infrastructure is not speculative. The alternative is the current state: each program generates data that benefits one person and then stops generating value. The data sits in a filing cabinet or a secure server at one institution. The next program at another institution starts from a higher baseline than milasen started from, because published case reports and regulatory precedents exist, but from a lower baseline than it could start from if structured, queryable data from all prior programs were available.

The difference between a published case report and a structured, queryable dataset is the difference between reading about someone else's experience and having access to their raw data. The case report says "the ASO was well tolerated." The dataset says "CSF pleocytosis of grade 1 occurred at day 3 post-injection, resolved by day 7, across 8 of 12 treated individuals, with no dose-limiting toxicity." The second enables dose optimization. The first does not.

The Human Meaning

A family whose child has been diagnosed with an ultra-rare neurological condition caused by a mutation no one else in any database shares faces a specific set of questions. Can an individualized therapy be designed? How long will it take? How much will it cost? What are the risks? What are the chances it will work?

The answers to every one of those questions improve with each person treated before their child. The safety database is deeper. The manufacturing process is more efficient. The regulatory pathway is more established. The dose-response relationship is better characterized. The predictive models for which mutations respond to which mechanisms are more accurate.

The family whose child is treated twentieth benefits from the data generated by the first nineteen. The family whose child is treated first generates the data that benefits the next nineteen. The asymmetry is real: the first family bears more risk and more uncertainty than the twentieth. The infrastructure that captures and shares data across programs is what makes the asymmetry shrink over time, so that the hundredth family bears meaningfully less risk than the first.

Each person treated is generating data that saves the next person. The only question is whether the data is captured in a form that makes the saving possible, or whether it disappears into institutional files and published abstracts that the next clinician has to reconstruct from scratch.

The drug is made for one. The data it generates serves everyone who follows.