ClearPath

Okay, so $QURE just got crushed nearly 40%, and a lot of people are reading that as the Huntington’s gene therapy falling apart at year four. I don’t think that’s what’s going on. What’s actually wild here is the comparison group. By year four, about half of those patients were gone from the data, and the ones who left were getting sicker faster. Leave them out and the remaining controls look healthier, so the drug looks weaker even when the treated patients themselves have not suddenly fallen apart. That pattern shows up all over medicine, and most people never see it. When a trial cannot ethically leave patients untreated for years, companies often borrow a “control” from a big natural-history database: people with the same disease, matched on paper, who were simply followed over time. It is a clever workaround. It is also fragile. The people who stay in those databases long enough to be counted at year four are not a random sample of the disease. The ones who get worse faster drop out, move to hospice, miss visits, or die. What remains is a quieter, healthier-looking average. Compare a real treated group to that thinned-out average and you can manufacture a “fade” that is partly just the control getting easier to beat. Huntington’s makes this especially sharp. As most of you know, it comes from one broken gene. That gene builds a sticky huntingtin protein that slowly wrecks the brain, and we still have nothing approved that puts the brakes on that decline. AMT-130, now named ifezuntirgene inilparvovec, tries something blunt and beautiful: one surgery, one dose into the striatum, the deep brain region this disease hits early, and a tiny designed RNA that tells cells to make less huntingtin. The high-dose group here is tiny, so every number needs humility. At three years, now with 15 people, the main disease score (thinking, movement, and day-to-day function rolled together) got about 80% less worse than matched people watched in that natural-history database. That three-year cut is what the company put in its filing with @FDA. At four years, 12 of those patients were still on that same score, and the slowing fell to 44%. It missed the usual statistical bar (p=0.144). The score that asks a simpler question, can you still work, manage a house, take care of yourself, still looked better: about 61% less decline (nominal p=0.008). That everyday-function score is also what the bigger confirmatory trial will lead with. And look at what happened underneath those percentages. By month 48, 53% of the matched comparison patients had missing follow-up. The ones who vanished were declining faster than the ones who stayed. If the sickest untreated people leave the spreadsheet, the remaining “no-drug” group looks healthier than real Huntington’s usually does. The gap between drug and no-drug then shrinks on paper. When the company ran the same four-year patients against an older version of that control set, the slowing on the main score came back near 54%. Same treated people. Different ghost of a control. Different story. To me the sell-off looks like traders bracing for a messier approval path, not like the biology flipped overnight. Twelve patients cannot settle how long a benefit lasts. What this episode can settle is a harder habit: before you trust a year-four “fade,” ask who is still left in the untreated column. What I am sure is that I truly hate all brain diseases. Disease in general is bad, yes. But brain diseases deserve our special hatred. We should unite as humanity to beat all brain diseases as quickly as possible. That I know...