For One

I

On 25 February 2025, at a hospital in Philadelphia, a six-month-old baby received by vein a few millilitres of a liquid made for him and for nobody else.

In the papers he appears as KJ. He had been born with a urea cycle disease: he lacked a liver enzyme [the cycle has six and one of his was failing] and could not get rid of the ammonia the body produces in processing proteins, so that the ammonia built up in his blood and damaged his brain.

Half the children with that severe form die in the first months. Those who survive usually reach a liver transplant with neurological damage.

What he was given was not an existing drug. It was an edit designed against his particular mutation —which he shared with almost nobody in the world— made from scratch in six months by a team from two hospitals.

He received three doses: the first in February, and two more in March and in April. He left the hospital and went on growing.

II

It has to be said what “bespoke” means, because it is the whole word of this piece.

Up to here, even the most sophisticated thing has been a product. A batch is made, it is tested in a group, it is approved for an indication and it is given to everybody who fits it. Even the living-cell treatment of the previous piece —which is made with one person's cells— always uses the same added piece.

Not this. Here what was designed was the correction of one particular letter of the genome of one particular patient, with a guide made to the measure of his sequence.

There is no batch. There is no indication. There are no other patients.

For that very reason there is no possible clinical trial in the classical sense; you cannot compare one group with another when the group is of one.

III

The only thing I can put here is a piece of paper I signed without reading.

It was at a hospital, in the blood-taking room, before an ordinary test. On the counter there was a two-sided form about the use of leftover samples for research, with two boxes at the end [yes and no] and a line for the signature.

I ticked yes. It took me four seconds [the pen was tied to the counter with a string].

I went back to find that document while writing this and read it whole for the first time: it says that whatever is left of my blood may be “kept anonymised”, used in future related studies, transferred to third parties under research agreements, and that I will receive neither compensation nor information about the results.

All of that is reasonable, is well drafted and is what allows the existence of the databases with which things like KJ's are designed.

My entire participation in the matter lasted four seconds and cost me nothing.

IV

Inside, this no longer cuts or pastes. It corrects a letter.

The first generation of gene editing worked like scissors. Both strands of the genetic material were cut at a chosen place and the cell was left to repair the cut, with the risk that it would repair badly or stick things where it should not.

What was used here is another thing. It is a tool carrying a finder and a chemical corrector. The finder is a guide strand, written to fit only the exact point of the genome where the spelling mistake is, and the corrector is an enzyme that chemically transforms one letter into another without cutting anything.

There are no scissors, no break and no emergency repair. There is one letter that changes.

The delivery is the part we already know from the previous piece, and the instruction goes inside a bubble of fat injected into a vein. Those bubbles, by their size and by their chemistry, end up above all in the liver [the default destination of almost everything injected wrapped in fat], and the liver was precisely where the problem was.

The liver cell receives the instruction, makes the tool for a few hours, the tool looks for the point, corrects the letter and afterwards everything that was injected degrades and disappears.

What is left is the change in the text, which is permanent and which the cell copies every time it divides. Only a proportion of the liver cells is corrected, not all, and a proportion is enough if that proportion makes sufficient enzyme.

V

What this case opens is not a therapy. It is an enormous and rather beautiful administrative problem.

The whole machinery of drug approval is built on the idea that there is a product and a population. It is tested in many, the average is measured, it is authorised and it is sold [that has been the order since 1938]. Here there is no population: there is one person, and the treatment made for him is of no use to anybody else.

What can be approved, then, is not the product but the procedure. The platform, the design method, the quality controls and the time it takes to assemble; and after that, each patient is a case.

That forces a remaking of the rules that came out of a poisoned syrup in 1937 and that we have used ever since.

It has to be distinguished what is remade and what is not. What cannot fall is the check that something does not poison, which cost a hundred and seven dead to learn. What can change is the idea that a population is needed to authorise a treatment, because there are diseases whose entire population is one family.

VI

The objection is the number and it is devastating.

A treatment like that cost, according to those responsible, something of the order of what a transplant costs, and it was done with public and academic funding, with no company behind it and with no intention of selling it. It could be done because the team was in the same place, because they had the platform set up and because the case was urgent and in the news.

Multiply it by the millions of people who have rare mutations and the sum does not work out by any means.

So this, today, is not personalised medicine. It is a demonstration that it can be done, done once, with a child who had the luck to be born twelve kilometres from the right laboratory.

Against me: I have written a great deal about the price going at the end and without comment, and here the price is the only thing there is.

VII

The paper I signed in four seconds has more to do with this than it seems.

Between that form and the liquid KJ was given there is a chain that starts in millions of anonymised samples, goes on through databases of variants, passes through prediction atlases and ends in a design made for one single person.

The first body of this work, in 1922, had a name and was fourteen and was in a room with other children. The last has no public name, is six months old, and his treatment exists because a very great many nameless people signed a paper like mine.

A hundred and three years between the two scenes.

And in both, what was administered had been made for one particular body that was dying. Collip's extract was purified over twelve days for a boy who had gone two years without eating; the liquid of February was designed over six months against the spelling mistake of one single child.

What changed in between was everything else.

11 August 2024, birth and diagnosis.

25 February 2025, first dose; March and April, the other two.

Six months from the diagnosis to the treatment [from 11 August to 25 February].

One letter of the genome corrected in a proportion of the liver cells.

Number of patients that medicine is of use to: one.

A sum that does not close: it is reckoned that there are between seven and ten thousand rare diseases described, most of them from point mutations like his, and that together they affect some three hundred million people [one in every twenty-five].

Time it took me to sign over my samples: four seconds.