The harness for n‑of‑1 medicine, for biotech and life sciences
Where biotech meets the patient first, in clinics running trial-grade protocols one patient at a time
Every patient at these clinics, on GLP‑1, hormone and regenerative protocols, is a clinical trial of one, and that is what opens partnerships with biotechs for decentralized trials, starting from a therapy these clinics already deliver, with baselines already on the record and facts that walk back to their source.
On this page7
What does this have to do with clinical trials?
iksa is the harness for n‑of‑1 medicine, built to run clinical trial grade medicine for individuals, and what a study gets from it is the loops you know from a trial, closely monitoring certain biomarker changes and possible side effects, already run by a doctor around one patient at a time.
Each patient is read against their own baseline, which is what we mean by a clinical trial of one, and a formal n‑of‑1 trial adds the planned, repeated crossover periods, often randomized, that make the patient their own control.
The plan on the record names the markers, dated retests and targets the doctor set, the way a trial’s schedule of assessments sits in its protocol, except that this one moves with the patient, tightening when a dose changes or a side effect appears.
Care stays care, run by the patient’s own doctor, and we design a study on top of it with you and the clinics that would run it. We call that partnerships with biotechs for decentralized trials, and the regulator’s words for the same ground are “decentralized elements” (FDA, September 2024 (opens in a new tab)) and randomized trials “integrated into routine clinical practice”, where FDA writes that leveraging established health care institutions and existing clinical expertise “can reduce startup times and speed up enrollment” (FDA draft guidance, September 2024 (opens in a new tab)).
We have been asked by colleagues to replicate our protocols at scale to initiate genomics and metabolics studies for model training, and the doctors doing research at their own sites and taking it back to biotech organizations want to work with us for their clinical studies.
What could a study here answer?
If you run the protocol properly, the evidence falls out of care, so every protocol produces the evidence regulators require and life sciences acquires, and what a study receives is the record care already keeps. By statute, FDA runs a program to evaluate the potential use of real-world evidence “to help to support the approval of a new indication for a drug approved under section 355(c)” and “to help to support or satisfy postapproval study requirements” (21 U.S.C. 355g (opens in a new tab)).
A first study here leads with a therapy these clinics already deliver or a new use of one, what FDA’s draft words as approved drugs “being studied for new indications, populations, routes of administration, or doses” (FDA draft guidance, September 2024 (opens in a new tab)). It can be observational, including as an external control for a single-arm trial, or interventional: a randomized trial run inside routine care.
It starts from a recognized condition and the measures the record already has a place for, from labs and body composition to VO2 max, grip strength and scored cognitive and sleep scales, so body-composition trajectories under GLP‑1 weight loss, or menopausal hormone therapy after FDA’s label changes (FDA, February 2026 (opens in a new tab)), are the kind of question that fits.
A new candidate can fit too, where FDA’s draft applies to trials of unapproved drugs “when the safety profile is sufficiently characterized and the drug is appropriate to be administered and managed in the setting of routine clinical practice” (FDA draft guidance, September 2024 (opens in a new tab)).
The classes of medicine these clinics already run set where the fit is: GLP‑1 and metabolic, hormone, regenerative and peptide-class protocols, women’s health and integrative oncology.
These classes run from approved medicines to compounds with no human trial data yet, and one US pathway is opening toward that far end: in the notice of a draft policy, HHS says it doesn’t intend to take enforcement action against pharmacies and physicians compounding six peptides (Federal Register, October 2026 (opens in a new tab)) on the policy’s conditions (Hall Render, November 2026 (opens in a new tab)).
The draft approves none of the six, and FDA’s reviewers had recommended against listing any of them for compounding (FDA, July 2026 (opens in a new tab)) and identified no clinical studies or human exposure data for three of them (FDA’s reviews, July 2026: first (opens in a new tab), second (opens in a new tab), third (opens in a new tab)), while FDA’s advisory committee voted for listing all six (Hall Render, November 2026 (opens in a new tab)). So patients are reaching this medicine ahead of any large trial, and a study that reads what was given and what followed, patient by patient, is how that evidence starts.
On the endpoint, when the field pooled 51 intervention studies, drugs moved epigenetic clocks more than lifestyle did, yet no clinically important difference has been defined for a clock (Nature Medicine, August 2026 (opens in a new tab)), so a clock that moves on your therapy reads as context and the endpoint to build on is function, on the measures above, beside the conventional markers.
Dense repeated measures in the same people show how much a marker varies within one person, the background any change on a therapy is read against, and FDA lists that variability among the factors for judging a biomarker test for a stated use in drug development (FDA draft guidance, December 2018 (opens in a new tab)). The record is built toward those repeated measures.
We hold our words to the grade of the evidence, and what clinics alone can show becomes a candidate for prospective testing, and your study picks it up there.
Who are the patients, and what can’t claims data see?
They are long-term patients at cash-pay clinics, on protocols that stack, followed through repeated labs, and they are self-selected and health-seeking, so any design starts from that.
Because they pay the clinic directly, much of what they are on stays out of the data a sponsor usually buys, and FDA writes that EHRs and claims “may not systematically capture” drugs “not reimbursed under health plans” (FDA, July 2024 (opens in a new tab)).
Before a study question is asked, a patient’s record already holds dated lab series, notes, treatment plans and body composition, with genetics and wearable history where they exist, and their first work-up is kept as its own program, so the baseline is there to find. Notes, visits and payments are kept as dated history too, so how long a patient stays in care can be counted from the record and not estimated.
Have a population in mind? Bring your inclusion and exclusion criteria.
Start a study conversationWhat would a study ask of the clinic?
A study adds work on top of care, so we would walk your protocol through with the clinic and mark which steps are already usual care and which need trial personnel trained on it, the line FDA draws when it says the activities local providers perform “should not differ from those that they are qualified to perform in clinical practice” (FDA, September 2024 (opens in a new tab)).
Your study would start from a clinic that keeps charting in its own system, and the clinic’s own notes, reports and files are read into a structured record.
You stay responsible for the quality of what each site generates, and each site’s investigator for the trial’s conduct there, delegated tasks included (ICH E6(R3), FDA final guidance, September 2025 (opens in a new tab)), so here iksa’s AI agents at the clinic draft and propose and a person confirms.
The treatment companion for patients is built to answer only a patient individually activated on record, to hand anything clinical to a person and to leave diagnosis, prescribing and protocol changes with the doctor, and each patient will come on one at a time, on the doctor’s go-ahead, after a person has reviewed that patient’s profile.
Would the data hold up?
Your evidence lead will start with the plan each patient was on, and the record is built to hold each protocol with its dates, the therapies, the markers monitored, the retests the doctor dated and the targets, read from the doctor’s own notes.
A data point walks back to its source, because each fact read from the notes and reports carries where it came from and its date, or says plainly that it has none. Any quotation a doctor is shown comes from the record, never from the AI answering the question, and one that can’t be found there is never shown.
Source notes and their attached files stay exactly as the clinic’s system returned them, every version the record has copied kept side by side with the time it was first seen.
Whoever reviews the study will ask who did what and when, and staff verdicts on flagged items, proposals from iksa’s agents, the messages sent out and changes to roles and permissions go into an audit trail that names who acted and when.
Each new lab value is checked against the text of its report, and a value that fails stays off the doctor’s table, is kept and counted, and the kind of failure goes to a person to rule on.
Lab values carry standard lab codes and units where they are mapped, beside the value and unit exactly as the lab printed them.
An adverse event a patient reports is built to reach a person, going by urgency to the nurse or doctor who takes it, with the patient’s words. A contraindication the doctor writes in a note is carried into the patient’s safety section with its source, and a critical value on a lab the record accepts raises an alert the doctor sees on opening the patient.
Here is one reading, followed back to where it came from.
Put this in front of your evidence lead, and bring what they ask to the first conversation.
Start a study conversationWhere is this going?
Long term, iksa becomes the Flatiron of personalized medicine, sitting between the clinics running the hardest protocols and the biotechs formulating what they run. We believe life sciences goes to these micro-ecosystems inevitably, because the consumer demand is just too strong, and the developers on the other side are already at work, with 4,487 gene, cell and RNA therapies in development from preclinical through pre-registration (ASGCT and Citeline, July 2026 (opens in a new tab)).
Decentralized trials are how it starts, because the parts a decentralized study leans on are built for care at the clinic: a remote patient channel, wearable data, timed check-ins, per-patient permissions and chart upload, and binding them to one study is what comes next.
Patients are already in care with a doctor they know, so a study conversation starts with the clinic’s own patients, and from there a sponsor’s eligibility question will become a count from the record itself, in aggregate with small cells suppressed, with no row leaving the clinic for that count.
What the doctor planned and what was billed as given are both on the record, including exposures standard drug codes don’t cover, and reconciling the two is how a study will know what each patient actually received. The adverse-event loop is what the record is built toward, each event linked to the exposure it followed and counted per patient over time, with a person reviewing it.
Study-standard export is where the record is going, produced for the study that asks. Every cross-patient question will be written first as the trial it would have been, a target trial emulation reported according to the TARGET Statement, and locked before any outcome is read. Every estimate will carry its error measured against negative-control outcomes it should not move (empirical calibration).
The vision is autonomous personalized medicine in dispersed micro-ecosystems globally operating at trial-grade.
How does a first study conversation start?
You bring the indication, the phase and who would be eligible, and we come back with the clinics that could run it and the markers they read, so you can judge quickly whether this earns a place in your feasibility work.
A study would run on de-identified data, with each patient’s consent, under ethics review and the terms written into each clinic’s agreement.
Bring your questions on how your eligibility criteria meet these patients, who the investigator is at each clinic, what participants pay for on study, how your product is stored and accounted for, what a clinic’s other therapies mean for conduct and inspection, how investigators’ interests are disclosed, recruitment and retention, the contract and monitor access, and access to individual-level data and source records.
You would be asking them of clinicians, drug developers and engineers in one team, Hopkins-incubated (opens in a new tab) at the Johns Hopkins Pava Center, with people who come from Novartis, Novo Nordisk, GSK, Pfizer, Cytel and Medidata.
Bring your indication, and we’ll bring the clinics that could run it
Tell us the phase and who would be eligible, and we’ll come back with the markers those clinics read.