Newsletter · · Ashutosh Agarwal

The FDA Came Back and It Came Back Hard - Biotech Pipeline: Gene/Cell, Neuro & Tools - Week of August 2, 2026

A synthesis of what biotech and clinical podcasts said about gene and cell therapy, neurology and life-science tools for the week ending August 2, 2026, including the FDA publicly disputing Capricor's own trial claim, Vertex's $10 billion Crinetics deal, and new detail on Biogen's tau drug from the London Alzheimer's meeting.

Biotech Pipeline: Gene/Cell, Neuro & Tools

Week of August 2, 2026: The FDA Came Back and It Came Back Hard

Last week we ended on a cliffhanger: a stack of FDA advisory-committee meetings was landing, and the open question was whether the agency's new leadership would prove stricter, looser, or simply unpredictable. This week we got the answer, and it was not the gentle one. In the same few days, the FDA publicly disputed a company's own claim that its big trial had succeeded, something veteran reporters said they'd never seen before, and told a second company its data package might not be good enough to approve at all. Meanwhile, away from Washington, the sector's newly cash-rich middle tier went shopping: Vertex struck the biggest deal in its history, and the immunology star Argenx did its first-ever acquisition. And at the big Alzheimer's meeting in London, the field kept chewing on a tau drug that both thrilled and confused everyone who looked at it. Here's the week.

TL;DR

  • The FDA disputed a company's own "we hit our goal" claim, a first that rattled the whole cell-and-gene world. Ahead of an advisory panel, the agency's briefing documents flat-out disagreed that Capricor's Duchenne muscular dystrophy cell therapy had met its main trial goal, pointing to changes made to the statistical plan after the blinded part of the study ended. "I can't think of a time that I've ever seen this," one biotech editor said. The agency was just as tough on Replimune's skin-cancer drug. After months of looking flexible, the FDA is playing hardball again. (BioSpace, Jul 29)
  • The mid-cap biotechs are now the buyers. Vertex agreed to buy Crinetics Pharmaceuticals for about $10 billion in cash, the largest acquisition it has ever done, and reports earnings Monday. Days earlier, Argenx did its first-ever deal, paying roughly $2.2 billion for Forte Biosciences. As one analyst put it, "the buy-side roster has grown." (BioCentury This Week, Jul 28; Telltales, Aug 2)
  • Biogen's tau drug keeps dividing the room. New detail from the London conference: the drug actually made existing brain tangles shrink on scans, thought impossible, and slowed decline by 26% at its best. But it missed its main goal, the lowest dose worked best, and doctors now worry there may be such a thing as removing too much tau. (JAMA Medical News, Jul 31)

What's new

The story of the week: the FDA turned on its own approval math. For a year, the running question in biotech has been what the reshuffled FDA would actually do when the cameras came on. This week it showed us. On BioSpace's weekly podcast, editors Heather McKenzie and Annalie Armstrong walked through briefing documents the agency released ahead of two advisory-committee meetings, and the tone was startling. For Capricor Therapeutics, whose cell therapy, deramiocel, is aimed at Duchenne muscular dystrophy, a fatal muscle-wasting disease in boys, the FDA did something the hosts had never witnessed: it disagreed that the company's Phase 3 trial had met its primary goal at all. The agency's complaint was technical but damning: after the rigorous, blinded part of the study ended, Capricor made changes to its "statistical analysis plan," the pre-agreed rulebook for how you score the result, during the open-label stretch where everyone knew they were getting the drug, and that generated more than one version of the answer. The plan the FDA reviewers picked apart was, they said, a draft the company only handed over after the agency asked to see every version. "This was pretty stunning. I can't think of a time that I've ever seen this," McKenzie said. Capricor's CEO Linda Marban told BioSpace she was surprised the meeting was even called, and said she trusts the panel will "see through the statistical analysis plans and not turn this into something mathematical" and instead judge whether the therapy actually works.

The whiplash is the point. As Armstrong noted, "We've written a number of articles the past couple of months about how the FDA has seemed to be a little bit more flexible, and now it's kind of seeming like last year again." This is the same drug whose first advisory meeting, back in 2024, was abruptly canceled before its initial rejection, part of the leadership chaos that saw the agency's top cell-and-gene reviewers pushed out. Now the meetings are back, but the agency is holding the hard line those departures were supposed to soften. (BioSpace, Jul 29)

The second target was Replimune, whose melanoma drug RP1 was up for its third try at approval. The FDA again said the data, from a single-arm study (one with no comparison group) called IGNITE testing RP1 alongside the immunotherapy Opdivo, was "not interpretable" and "potentially not sufficient to support approval." Replimune's rebuttal is the one every small oncology company now reaches for: a randomized trial would have been unethical in these very sick patients, and rivals like Iovance, Merck and Bristol-Myers Squibb have won melanoma approvals on single-arm data. STAT's crew on The Readout Loud added the backstory: this drug has been rejected twice already, and the White House reportedly pushed the FDA to look a third time. (The Readout Loud, Jul 30)

The second story: the sector's rising stars have become the acquirers. The most investable theme of the week wasn't a data readout. It was a change in who writes the checks. On BioCentury This Week, the editors dug into Argenx's first-ever acquisition: it's paying $77 a share in cash, about $2.2 billion, for Forte Biosciences and its antibody against a target called CD122, which has shown early but promising data in the skin condition vitiligo and in celiac disease, and is being pushed into the hair-loss disorder alopecia. The logic is a copy-paste of Argenx's own success: take one antibody and expand it across many diseases, exactly what it did with its blockbuster Vivgart (which now has four approvals and did $1.5 billion in a single quarter, with a goal of ten approvals by 2030). Argenx can afford this because it's worth more than $50 billion and trading near all-time highs. As BioCentury's Paul Bananos put it, "the buy-side roster has grown," a new class of large-but-not-mega biotechs is now competing with big pharma for deals. His examples: Genmab paid $8 billion for Merus last year, Incyte just bought Vega for more than a billion up front, and, the headline, "Vertex just made a $10 billion deal." (BioCentury This Week, Jul 28)

That Vertex deal is the one our coverage list cares about most. On the Telltales weekend markets recap, the terms were laid out: Vertex, the Casgevy gene-editing partner and the biggest name in cystic fibrosis, agreed to acquire Crinetics Pharmaceuticals for roughly $10 billion in cash, about $85 a share, the largest acquisition in Vertex's history. It also signed a smaller, science-stage collaboration with AbCellera on next-generation "T-cell engagers," antibodies that grab a patient's own immune cells and point them at disease, for $28 million up front. Vertex reports earnings Monday (Aug 3), and as Telltales' hosts noted, the print "won't tell you anything about the acquisition. It'll tell you whether the base business is still carrying the load," because the existing franchise is funding the whole thing, and a deal this size is unproven integration for a company that has never bought anything close to it. (Telltales, Aug 2)

The tau drug keeps reverberating, and the closer you look, the odder it gets. JAMA's medical-news podcast sat down with Dr. Gil Rabinovich, who runs the Alzheimer's center at UC San Francisco and just got back from the Alzheimer's Association International Conference in London. The drug everyone was arguing about is Biogen's Duranderson, an "antisense" medicine that quietly reduces the body's raw material for making tau, the toxic protein that tangles up inside brain cells (the approved drugs all target amyloid, the gunk between cells). Rabinovich called it "a program that I'm actually very excited about," and the biomarker results are genuinely eye-popping: the drug lowered tau in spinal fluid by up to 60%, and, most striking, tangles that had already formed and lit up on brain scans "seemed to disappear or at least reduce in intensity." As he put it, "We thought the tangles were there to stay, but these results suggest that if you lower tau, they may be more dynamic."

So why is the field split? Because the Phase 2 study (called CELIA, roughly 400 early-stage patients) missed its main goal, which was to show that higher doses help more. The opposite happened. The low dose delivered the biggest cognitive benefit, about a 26% slowing of decline on the standard scale, while the mid and high doses came in at just 14% and 9%. And there was a dose-dependent side effect the company said little about: a "confusional state" that got more common the higher the dose. Rabinovich floated the unsettling explanation: tau isn't purely a villain, it does real jobs holding brain cells' scaffolding together, so "you can lower tau too much," and there may be a sweet spot above which scrubbing out more tau starts doing harm. One more logistical bright spot: the drug is delivered by spinal injection, but in the winning low-dose arm only every six months. (JAMA Medical News, Jul 31)

The quiet advance nobody put a stock ticker on: editing genes to lower cholesterol. On the MedEvidence podcast, two clinical-trial doctors in Jacksonville gave the clearest plain-English tour yet of in-body gene editing for heart disease, the frontier that Verve Therapeutics is built around. The idea: permanently switch off genes like PCSK9 (people born without it live normal lives, minus the heart attacks) or ANGPTL3, using a one-time infusion that travels to the liver, instead of a pill or an injection every few months. They're running one such trial now, and the early safety read is reassuring: the first roughly 30 patients (dosed in New Zealand, the UK and Australia) were published in the New England Journal of Medicine at the end of May with "no deaths" and "no major complications," and about 50 people worldwide have now been dosed. The honest caveat the doctors kept returning to is irreversibility: "one and done" is wonderful until you change your mind, and there's no undo. They also flagged the competitive backdrop: Merck's new oral PCSK9 pill, enlicitide, was just approved, so the bar for simply lowering cholesterol keeps rising on the cheap-and-easy end even as the gene editors chase the permanent fix. (MedEvidence! Truth Behind the Data, Jul 29)

And the reminder of why gene editing's commercial ramp is slow: the chemo. A sickle-cell physician on the Cykiert Files podcast gave a wonderfully clear account of how Casgevy, Vertex and CRISPR Therapeutics' approved editing therapy, actually works: it disrupts a genetic switch (an enhancer on a gene called BCL11A) so red blood cells restart making fetal hemoglobin, which climbs to 40–45% and lets patients "behave like somebody who has sickle cell trait." More than 90% of treated patients end up free of sickle-cell symptoms. But here's the part investors underestimate: before you can put the edited cells back, you have to wipe out the patient's bone marrow with busulfan, a chemotherapy that "has got pulmonary toxicity... but more importantly, it also causes infertility." That toxic, fertility-threatening conditioning step, not the edit, is a big reason a one-time cure still gets adopted slowly. (DoctorPodcasts, Jul 30)

The debate

Is the FDA's hard turn a healthy correction or dangerous whiplash? This is the argument the week actually staged, and both sides showed up on the same podcast. The case for the agency: a drugmaker rewriting its scoring rules after unblinding is exactly what advisory panels exist to catch, and a regulator that waves through weak single-arm data trains every company to bring less. The case against: the FDA is being accused of moving the goalposts after the fact and reviving the very rigidity that drove out its own experts last year, and, as BioSpace noted, some former regulators are puzzled the agency even "resumed adcomms at all while they're still trying to staff back up." For Capricor, the human stakes are Duchenne boys with no good options; for Replimune, 22 clinical leaders had petitioned in the drug's favor. The unresolved tension: is this a principled agency rebuilding its standards, or an unpredictable one whose answer depends on who's in the room that day? The votes this week will color how every cell-and-gene developer models its own path. (BioSpace, Jul 29; The Readout Loud, Jul 30)

One-and-done editing vs. the drug you can stop taking. The cholesterol discussion surfaced the field's other real tension, honestly argued by the doctors running the trials. The bull case for editing: chronic cholesterol drugs, even the newer injectables whose effect lasts four to six months, depend on patients remembering, insurers paying, and access never lapsing; a single permanent edit removes all of that. The bear case, in their own words: irreversibility cuts both ways, and with good, cheap options already on the shelf (now including an oral PCSK9 pill), a patient "should make the choice about getting these new options after that thoughtful process." Their rule of thumb, if you're doing fine on existing drugs, keep taking them, and reserve the permanent fix for people the current toolkit fails, is a useful reminder that the addressable market for one-and-done editing is narrower than the excitement implies. No one voiced the opposite extreme this week. (MedEvidence! Truth Behind the Data, Jul 29)

Read-throughs & names in play

  • Vertex (VRTX). The week's most consequential name for this list. Bull: a $10 billion, $85-a-share deal for Crinetics is the biggest bet in company history and shows the cystic-fibrosis cash machine is being redeployed to diversify beyond CF; the AbCellera T-cell-engager tie-up adds a cheap option on a hot autoimmune modality. Bear: it's an unproven integration funded entirely by the existing franchise, and Monday's earnings will reveal whether that base business is still "carrying the load." Watch the print for CF and Casgevy momentum, and remember the busulfan-conditioning drag on Casgevy uptake. Next catalyst: Q2 earnings, Monday Aug 3. (Telltales, Aug 2; BioCentury This Week, Jul 28; DoctorPodcasts, Jul 30)
  • The Duchenne complex (Capricor, Sarepta). Capricor is the acute risk: an FDA that won't concede its Phase 3 hit its endpoint is an existential problem, and the adcomm vote is the binary. Read-through to Sarepta, which just named a new CEO, Michael Severino, previously head of the gene-editing startup Tessera Therapeutics, as it works through its own DMD safety and efficacy scrutiny. A hard FDA on Duchenne cell therapy raises the bar for everyone in the muscle-disease space. (BioSpace, Jul 29; The Readout Loud, Jul 30)
  • Verve and the in-vivo cholesterol editors. No company-named data this week, but the MedEvidence trial tour is squarely supportive of the thesis: an in-body PCSK9 editing program with ~30 patients now in the NEJM and no deaths so far, and ~50 dosed globally, means the safety base is quietly widening. The offsetting read: Merck's newly approved oral PCSK9 pill (enlicitide) keeps raising the "why not just take a pill" bar the editors must clear on convenience and durability. (MedEvidence! Truth Behind the Data, Jul 29)
  • The ATTR editing read-through (Intellia, CRISPR Therapeutics). Indirect but worth logging: AstraZeneca said its ATTR drug Wainua failed in the heart form of the disease (ATTR-CM), a stumble for the RNA-silencing approach that competes with the one-time gene-editing programs aimed at permanently switching off the TTR gene. When the chronic silencers wobble, the "edit it once" pitch gets relatively more attractive. (BioSpace, Jul 29)
  • Biogen (BIIB), the tau overhang continues. Bull: first drug ever to shrink existing tau tangles on a scan, with low-dose cognitive slowing (26%) in the range of approved amyloid drugs, and only twice-a-year dosing. Bear: missed its primary goal, a backwards dose-response, a dose-dependent confusion side effect, and a live worry that lowering tau too far could backfire. The whole tau field, anyone chasing it with antisense, siRNA or degraders, inherits that "is there a ceiling?" question. (JAMA Medical News, Jul 31)
  • The anti-amyloid launch, from the patient's chair. A rare on-the-ground look at how Leqembi (lecanemab) actually lands: Scott Redfern, diagnosed with younger-onset Alzheimer's at 61, described seven months of biweekly infusions with "no effects at all," no ARIA (the brain-swelling risk), no headaches, after clearing the full gauntlet the launch depends on: a blood p-tau217 test (which came back inconclusive for him), an amyloid PET scan (which confirmed it), and genetic testing (he carries no high-risk APOE4 copies). He chose lecanemab over Lilly's Kisunla for its longer-term plaque-clearing. It's a reminder that the diagnostic plumbing, and patient self-advocacy, gates the whole anti-amyloid market. (Brain Talk | Being Patient, Jul 28)
  • The blood-test build-out (read-through to the whole neuro launch). From the AAIC recap: in a study of ~2,700 cognitively normal older adults, those with high p-tau217 at baseline had a 24% chance of cognitive impairment within five years, rising to 38% for the "very high" group, powerful for enriching trials, but the updated clinical guideline still recommends against testing healthy people outside research. Bigger picture, Rabinovich argued the future is combination therapy (amyloid + tau at once) and pointed to a new NIH-funded trial at UCSF doing exactly that, while stressing "the tremendous importance of continued federal funding," a quiet flag for any life-science tools or diagnostics name levered to NIH budgets. (JAMA Medical News, Jul 31)
  • The deal machine keeps annexing gene editing. On RARECast, Chiesi's rare-disease head Giacomo Chiesi detailed a CRISPR-Cas editing collaboration with Arbor Biotechnologies (struck last October) aimed at the kidney disorder Primary Hyperoxaluria Type 1, one of four modalities the company is assembling, alongside a $1.9 billion buy of Calvista. Read-through: mid-size pharma is steadily licensing editing platforms rather than building them, which keeps demand under the tool-and-enzyme suppliers even when the marquee editing stocks are quiet. (RARECast, Jul 30)

What changed

Last week we asked whether the returning FDA would be stricter, looser, or unpredictable. This week answered: stricter, strikingly so, to the point of contesting a company's own claim that its pivotal trial worked. That single fact reprices risk for every cell-and-gene developer counting on a friendlier agency. The second shift is in the M&A map: the buyers are no longer only big pharma. A tier of large-but-not-mega biotechs, Vertex, Argenx, Genmab, Incyte, now has the cash flow to write nine- and ten-figure checks, which changes the exit math for every private developer and every takeout target. The tau story didn't so much change as deepen: we now have the specific, awkward numbers (tangles shrinking, but a missed endpoint, a backwards dose curve, and a real fear of over-lowering tau). And the cholesterol-editing safety base widened another notch without a headline.