Two Children Got Drugs Made Only for Them. One Walked for the First Time at 15.

$200,000.

That is the approximate cost to design, manufacture, and file a regulatory application for a drug that will only ever treat one human being on Earth. In the world of pharmaceutical development, where the average approved therapy consumes $2.6 billion and 10 to 15 years before reaching a patient, that figure is so small it barely registers as a rounding error on Pfizer's quarterly R&D budget. But for a 14-year-old boy with SCN2A-related developmental epileptic encephalopathy who had never walked independently, it was the difference between a wheelchair and standing on his own two feet at age 15.

A paper published July 21 in Nature Medicine describes the two-year results of two separate n-of-1 clinical trials conducted at UC San Diego and Rady Children's Institute for Genomic Medicine. Each trial enrolled exactly one patient. Each patient received a custom-designed antisense oligonucleotide: a short synthetic piece of DNA engineered to silence the specific mutant copy of their SCN2A gene while leaving the functional copy intact. Nothing about the results is subtle.

By the Numbers

Patient One was nine years old at trial enrollment and seizing nearly every day, a brutal rhythm that had resisted every medication thrown at it. Over two years of intrathecal ASO treatment, administered every two to three months via lumbar injection under anesthesia, seizure frequency dropped 26%. That sounds modest until you account for the context: this child had already failed multiple anti-seizure drugs, and any measurable reduction in truly refractory epilepsy represents clinical ground that years of conventional pharmacology could not take no matter how many approved medications were tried.

Patient Two is the headline, and the number is ninety percent. That is how far his seizure frequency dropped at age 14, eventually reaching stretches of entirely seizure-free days his family had never witnessed. He cut back on anti-seizure medications. His chronic gastrointestinal problems eased. Then the part that stops you cold: he walked on his own for the first time in his life at 15.

He walked.

Both children showed improvements in language and motor skills, sensory processing, adaptive behaviors, and a reduction in autism-related behaviors, while neither experienced serious side effects and routine bloodwork, ECGs, and EEGs remained stable across the full two years of monitoring that these trials demanded before the investigators would draw even preliminary conclusions about what the therapy could do.

How a Drug Gets Made for One Person

The mechanism is elegant and surprisingly simple. SCN2A mutations are typically de novo: spontaneous, not inherited. They corrupt one of the gene's two copies, producing a defective sodium channel protein that drives abnormal brain excitability. But diploid genetics offers an exploit because everyone carries two copies of every gene, and in these patients, one copy works fine.

Rady's team sequenced each child's whole genome, identified the pathogenic mutation, then looked for benign genetic variants: harmless single-nucleotide polymorphisms sitting near the mutation on the same chromosome. They designed an ASO that binds those benign markers, not the mutation itself, which means the ASO exclusively targets the chromosome carrying the broken gene. Once bound, the mutant gene is silenced while its healthy counterpart keeps working and sodium channels normalize.

"The therapy is deliberately designed to target the individual's genetic diagnosis," said principal investigator Olivia Kim-McManus. "The ASO modifies genetic expression and what proteins are expressed." This is not gene editing, and nothing is permanently altered, because the ASO degrades over weeks, which is why repeat dosing is required, but it also means the therapy is inherently reversible if anything goes wrong or the patient's condition changes in ways that make continued treatment inadvisable. Stop dosing. Effects wane.

A Pipeline That Doesn't Exist

Pharma runs on scale: spend billions developing a drug that treats millions, then recoup costs through volume over a patent's lifetime. An N-of-1 drug inverts that logic completely, because the entire market is a single patient and the "pipeline" is a bespoke manufacturing run measured in milligrams rather than metric tons.

We can reconstruct the approximate cost structure: whole genome sequencing runs $300 to $1,000; computational ASO design and in vitro screening costs $20,000 to $50,000; GMP manufacturing of a custom oligonucleotide batch runs $50,000 to $100,000; abbreviated preclinical safety studies add $30,000 to $80,000; and FDA research IND filing costs $10,000 to $20,000. All told, development cost falls somewhere between $110,000 and $250,000 per patient.

Ongoing treatment runs cheaper still. ASO synthesis costs $500 to $2,000 per dose. Each intrathecal injection procedure (lumbar puncture under anesthesia) runs $3,000 to $8,000 at a hospital facility. At four to six doses per year, annual treatment cost lands between $14,000 and $60,000. Compare that to the annual direct medical cost of refractory pediatric epilepsy: $16,000 to $47,000 in medications, emergency visits, and hospitalizations alone, according to CDC data, before counting neurodevelopmental support services that can add another $10,000 to $30,000 per year. A lifetime of uncontrolled seizures in a child carries an estimated direct cost of $1 million to $4 million.

Read that again. The personalized ASO may not just be more effective than the status quo but also cheaper to maintain over a child's lifetime, which inverts the assumption that personalized medicine must always cost more than the one-size-fits-all approach it replaces.

Cost Category	Traditional Drug Dev.	N-of-1 ASO
Development cost	$2.6 billion (avg.)	$110K–$250K
Timeline to first patient dose	10–15 years	~12 months
Patients served	Millions (if approved)	One
Annual treatment cost	Varies widely	$14K–$60K
Reversibility	Drug-dependent	Inherently reversible

Who Else Could This Help?

This is where the math gets interesting, and where nobody else has run the numbers. A JAMA Neurology paper published earlier this year assessed 160 infants with genetic epilepsies across 172 variants and asked a straightforward question: how many could theoretically be treated with ASO therapy? Sixteen percent. Twenty-five out of 160 infants had variants amenable to ASO approaches: 54% via gene knockdown, 29% via wild-type upregulation, 13% via splice correction, and 4% via exon-skipping. Of those 25, 68% could be considered for therapy right now.

Now scale that up to the national population. Approximately 3.4 million Americans live with epilepsy. Roughly 30% (about a million people) have drug-resistant seizures. Among pediatric epilepsy cases, genetic causes account for an estimated 30 to 50% of cases. If we conservatively apply the 16% ASO-amenability rate from the JAMA study to the subset with identified genetic causes, and assume a pediatric population of roughly 470,000 children with epilepsy (the Epilepsy Foundation's estimate), we get a rough calculation:

470,000 children × 30% drug-resistant × 40% genetic cause × 16% ASO-amenable = approximately 9,000 children in the United States alone who might benefit from personalized ASO therapy. That number does not justify a blockbuster drug pipeline, but it is not zero either. Nine thousand children seizing through medications that do not work, some of whom might walk for the first time if someone designed a drug just for them. That number haunts.

Full-Strength Counterargument

Two patients, open-label, no placebo arm, no blinding — every limitation you can name is here. Each child served as their own control, which means we are comparing a child's seizure frequency before treatment to their seizure frequency after treatment, with no way to separate the ASO's effect from the natural history of the disease, regression to the mean, observer bias, or the developmental gains that children make over two years simply by growing. The 26% seizure reduction in Patient One is within the range achievable with some standard anti-seizure medications; the improvement could be coincidental.

Patient Two's 90% reduction is harder to dismiss, and so is walking independently for the first time, but "harder to dismiss" is not the same as "proven" when the open-label design means every clinician administering the drug, every parent counting seizures, and every therapist measuring developmental milestones knew the child was on an experimental therapy. Expectation effects in pediatric neurology are large and documented. Bias lurks.

Scalability poses a harder question, and it is the one that separates an inspiring clinical anecdote from a viable therapeutic platform that could reach the thousands of children who might benefit from this approach. Designing a custom ASO for each patient requires whole genome sequencing, bioinformatics analysis, SNP identification, ASO design and screening, GMP synthesis, preclinical testing, and an FDA IND filing. For two patients at a well-resourced academic medical center with a dedicated genomic medicine institute, world-class sequencing infrastructure, and an established relationship with the FDA's research IND office, that pipeline is feasible. For 9,000? Nobody knows. No existing regulatory infrastructure supports this at scale, and FDA's research IND pathway was not designed for thousands of individual applications per year. Manufacturing capacity for GMP oligonucleotides is finite, and expert labor (genomic scientists, ASO designers, regulatory specialists) cannot be cloned any faster than the ASOs can be synthesized.

What We Don't Know

This analysis relies on published results from two patients followed for two years, and every projection built on that foundation should be held with appropriate skepticism. Whether seizure reduction persists beyond that window is unknown. Whether developmental gains continue, plateau, or reverse with long-term dosing is unknown. JAMA's amenability estimates are based on variant classification, not clinical outcomes. That gap between "theoretically amenable" and "clinically effective" is where most therapies die. Our cost estimates use published literature on ASO manufacturing and hospital procedure pricing, but actual costs in a clinical setting may differ substantially based on institution, insurance negotiation, and GMP facility overhead. Addressable population calculations assume genetic testing penetration rates and diagnostic accuracy that may not reflect current clinical practice, where many children with epilepsy never receive genetic testing at all.

The Bottom Line

For decades, the pharmaceutical industry has optimized a model where every drug must serve millions to justify its existence. This study asks whether the model can be inverted: whether a drug designed for a single person, developed in months instead of decades, at a cost three to four orders of magnitude below the industry average, can produce outcomes that conventional medicine cannot. Two patients is not proof. It cannot be. But a 15-year-old walking for the first time is not noise, either.

Practical Takeaways

If you have a child with drug-resistant epilepsy who has not had genetic testing, ask your neurologist about whole-exome or whole-genome sequencing. Sequencing costs have dropped below $1,000, many insurance plans now cover it for refractory cases, and the JAMA Neurology study suggests that roughly one in six children with genetic epilepsy may have a variant amenable to ASO therapy, which means access to this intervention begins with a test that costs less than a new iPhone. You cannot access a treatment that requires a genetic diagnosis without the diagnosis.

If you are a researcher or clinician: the Kim-McManus paper in Nature Medicine includes methodological detail on the allele-selective ASO design process and the FDA research IND pathway for n-of-1 trials. Their regulatory playbook is replicable, and the authors explicitly call for expanding the model beyond SCN2A to other monogenic neurological conditions. Nothing here is locked behind a proprietary wall. Every method is published, and what remains is building the infrastructure to do it at scale.

For policymakers: the FDA's research IND pathway works for a handful of patients but not for thousands. If N-of-1 therapies are going to move from academic curiosities to standard of care, someone needs to build a regulatory framework that can process individualized drug applications at volume without drowning in paperwork. Rady Children's team proved the science works, and now the bottleneck is institutional.