Guérir Quelquefois

Carved into the tradition of the sanatorium at Saranac Lake, and attributed variously to an anonymous medieval Frenchman and to Edward Livingston Trudeau himself, is a sentence that has survived every advance in medicine since: Guérir quelquefois, soulager souvent, consoler toujours. To cure sometimes, to relieve often, to comfort always. Observe the architecture of the thing. Ambition descends as frequency rises; the rarest achievement is named first and immediately qualified. It is a physician's aphorism written by someone who had watched a great many patients die.

The modern biotechnology deck has inverted the sequence. Cure is now the most frequently deployed word in the sector and the least frequently delivered thing in it. This would be merely tiresome, the ordinary inflation of commercial speech, were it not that the word is also load-bearing. It holds up prices. It holds up models. It holds up, in several notable cases, an entire capitalization table.

So it is worth asking a question that sounds like a schoolman's diversion and is nothing of the kind: is a cure a prevention or a treatment?

The reflexive answer is that it is obviously a treatment. One cures what already ails. Prophylaxis acts before the pathology; therapy acts upon it; a cure is simply therapy that finishes the job. Clean enough. Now press on it. Sofosbuvir cures hepatitis C, and in curing it prevents cirrhosis, hepatocellular carcinoma, and a place on the transplant list. Was that a treatment for a virus or the prevention of a cancer? It was both, in sequence, in the same patient, from the same pill. The category one assigns depends entirely on where along the natural history of the disease one chooses to stand, and that choice is made not by the molecule but by the person describing it.

This is not lexicography. Categories carry prices. Prevention is sold to the well: an enormous population, a thin per-capita margin, procurement by ministries and public programs, and a permanent flavor of political scrutiny. Treatment is sold to the sick: a narrow population, an urgent buyer, and a willingness to pay that scales with fear. The same intervention, filed under different headings, produces different profit-and-loss statements. A firm that can plausibly describe its asset either way will describe it whichever way the room requires, and will not think of itself as lying.

Then there is the deeper structural embarrassment, which the industry has never resolved and mostly declines to discuss in daylight. A chronic treatment is an annuity. A cure is a liquidation. Gilead's hepatitis C franchise took in something on the order of nineteen billion dollars in 2015 and less than a fifth of that three years later, not because the drugs failed but because they worked: the prevalent pool was drained, and what remained was the far smaller flow of incident cases. In April 2018 an analyst at Goldman Sachs put the matter in a report title that was received as a scandal (whether curing patients is a sustainable business model) when it was in fact merely arithmetic. A cure retires its own customers. Every honest model in the sector already knew this. The offense was writing it down.

Which brings us to the one-time genetic therapies and their remarkable prices. Zolgensma, Hemgenix, Casgevy, Lyfgenia: numbers in the millions, defended explicitly against the lifetime cost of the alternative. Note what is doing the work in that defense. The word cure is not ornamental there; it is the premise of the discounted stream. Remove it (concede that the effect may attenuate at year eight, that the vector may not persist, that the patient may return to standard of care) and the number becomes indefensible in an afternoon. This is why one increasingly finds outcomes-based warranties attached to such products, the manufacturer agreeing to refund the payer should the cure uncure itself. The warranty is an admission. One does not insure against an outcome one believes to be impossible.

Now the second half of the question, and the useful half: does the answer change depending on which representative of the company one asks?

It does, reliably, and the pattern is legible enough to be used as an instrument.

The chief medical officer will not say cure. She will say durable response, sustained remission, potentially transformative, and she will attach a median follow-up measured in months. She is speaking to the label, and the label is adjudicated by people who require evidence across a horizon nobody has yet observed.

The chief financial officer will say one-time, potentially curative. That adverb is the entire prospectus. He is modeling a bolus of prevalent patients over three to five years followed by a thin incident-only steady state, and he is hoping the question about year six does not come up before the raise closes.

The chief executive, onstage in San Francisco in January, will say cure, unhedged, to applause.

And market access will say whatever the payer needs to hear, which is a different sentence in every jurisdiction.

Meanwhile the euphemism of the moment, functional cure, spreads through hepatitis B and HIV: an adjective that quietly empties its noun. It means the disease remains and has agreed to behave. As a clinical concept it is respectable and even hopeful. As an investment concept it is a trapdoor with a rug over it.

The practical instruction follows. Ask the question three times, of three people, at the same company, and record not the answers but the variance among them. The variance is the finding. Modest divergence is the ordinary drift of a word conscripted into too many jobs at once, and should be forgiven. Wide divergence means one of two things: either the company does not understand what it owns, or it understands perfectly and has assigned a different truth to each audience. The first is incompetence and the second is something worse, and the market has only one way of pricing either.

Trudeau's sentence had the singular virtue of honesty about its own limits. Sometimes. The modern deck promises always; pressed, it means functionally; pressed again, it means for as long as anyone has bothered to look. The corruption of the language precedes the corruption of the thought, as a more famous essayist observed, and in this sector it precedes the corruption of the balance sheet by roughly one funding round.

So when the check is in front of you, hold the modern Guérir Quelquefois up against the old one. The medieval version graded itself honestly, ascending from the rarest thing to the most reliable one and admitting at the outset that the first was uncommon. The version being sold to you runs in the opposite direction. It promises always from the stage and discloses its sometimes in the extension study, in a footnote, eighteen months after the raise. The entire equity sits in the gap between those two documents.

And notice that the drift is not only between firms. It is inside them, frequently inside one person on the same afternoon. The chief executive who says cure in the morning will say durable response to a regulator after lunch and will mean no dishonesty by either, because the word has been quietly allowed to carry a different load in each room. That is exactly what makes it worthless as evidence and valuable as a signal. Price the variance, not the adjective. A company whose senior people converge, unprompted and across audiences, on the same careful sentence has decided what it owns. A company whose sentence changes with the room has not, and you are being invited to underwrite the difference.

The right question, then, was never whether the drug cures. It is what the word has been hired to do in the sentence, and who is paying its wages.