Editor’s Briefing: September 8, 2026

In Short. A week that taught the same lesson twice: moving a blood marker is cheaper than moving a patient’s future, and regulators will sometimes bless a new way of deciding when to switch therapy before the scan catches up.

---

The deal that needed a primary endpoint

Novartis spent about $12 billion to buy Avidity Biosciences and put an antibody-oligonucleotide conjugate, del-desiran, at the center of that bet. On Tuesday the company said the phase 3 Harbor trial in myotonic dystrophy type 1 missed its primary endpoint. In 159 patients, seven infusions given every eight weeks did not beat placebo on video hand opening time, the timed measure of delayed muscle relaxation that defines much of the day for people with DM1. (Fierce Biotech and BioPharma Dive)

The company still reports signals on secondary endpoints and exploratory analyses, and it says it will talk with health authorities about what comes next. Two other Avidity neuromuscular programs remain in the bag, including a Duchenne filing path and FSHD follow-up talks. The Harbor miss still lands where acquisition theses land when the largest peak-sales slide was the one that failed first. Shares fell hard in early trading, and a rival working a related DM1 approach moved with them. For a reader who cares about measurement, the teachable object is the endpoint itself: a functional timed test with high day-to-day noise can fail a program even when biologists still see “activity” elsewhere in the table.

Biomarkers that move, outcomes that do not

Four days earlier, Novartis and Ionis delivered a cleaner version of the same warning in cardiovascular disease. Pelacarsen, an RNA medicine built to plunge lipoprotein(a), failed a multiyear phase 3 study of more than 8,000 people on the hard outcomes that matter for heart programs: heart attack, stroke, and death. Lp(a) still fell. The events did not. (BioPharma Dive)

That pattern is older than this week, and it keeps returning because epidemiology and genetics made Lp(a) look causal long before anyone had a late-stage drug. The Harbor and Horizon pair belong in the same notebook. A count that moves on a lab slip can still leave the patient-facing claim empty. Programs that treat the biomarker as the finish line will keep meeting that wall.

When the FDA blesses a switch before the scan

Against that backdrop, AstraZeneca’s accelerated approval of Etcamah (camizestrant) is the week’s constructive surprise. The oral SERD won a label, with a CDK4/6 inhibitor, for HR-positive, HER2-negative advanced breast cancer when an ESR1 mutation appears during first-line aromatase inhibitor plus CDK4/6 therapy. The decision followed a hostile advisory committee and a review extension. Serena-6 had enrolled people whose scans still showed stable disease while circulating tumor DNA already carried the resistance mutation. Switching early cut the risk of progression or death by 56 percent, with median progression-free survival of 16 months versus 9.2 months for patients who stayed on the original regimen. (Fierce Biotech)

Guardant360 CDx received a matching FDA nod the same day so clinicians can retest about every three months and catch ESR1 emergence before radiographic progression. (Fierce Biotech) The approval is accelerated, so confirmatory work remains, and Serena-4 still has to read out. Even so, the agency has now endorsed a concrete claim: liquid biopsy can redefine the moment of intervention. That is a measurement story with a therapy attached, which is rarer than a therapy story with a biomarker decoration.

Lungs, once a day

At the European Respiratory Society meeting in Barcelona, Roivant’s Pulmovant unit reported phase 2 PHocus results for mosliciguat, an inhaled soluble guanylate cyclase activator licensed from Bayer for a modest upfront fee and larger biobucks. In 135 people with pulmonary hypertension tied to interstitial lung disease, the drug delivered a placebo-adjusted 56.3 percent drop in pulmonary vascular resistance at week 16, plus gains on six-minute walk distance and NT-proBNP, with a cough rate that did not exceed placebo. The company has already started phase 3 and is aiming at a large untreated population where Tyvaso has been the main branded inhaled option. (Fierce Biotech and BioPharma Dive)

AstraZeneca used the same congress to thicken the case for tozorakimab, its IL-33 antibody in COPD. Detailed Oberon and Titania data showed roughly 30 percent reductions in moderate and severe exacerbations across broad populations, with signals spanning eosinophil strata, and the program now carries FDA priority review. Barr called the molecule an internal science success story after the company redesigned it to block both arms of IL-33 signaling. (Fierce Biotech)

Rare disease gravity and a calendar that still ticks

Ultragenyx’s phase 3 Aspire study of apazunersen (GTX-102) in Angelman syndrome missed both its primary Bayley-4 cognitive endpoint and its multidomain secondary, sending the stock sharply lower and forcing a review of the program plus talk of significant expense cuts. (BioPharma Dive) Novo Nordisk, meanwhile, ended two more phase 3 trials of ziltivekimab in heart failure after a data monitoring committee saw little chance of escaping the logic of Zeus, the earlier ASCVD and kidney miss. Artemis after acute heart attack continues. (Fierce Biotech)

Elsewhere in Tuesday’s tape, Bristol Myers said a GPRC5D-directed CAR-T, arlo-cel, hit its primary response endpoint in heavily pretreated myeloma, and Pharvaris reported an 83 percent cut in hereditary angioedema attack rates in a phase 3 preventive study. (BioPharma Dive) Those are real clinical moves. They sit beside the week’s larger argument about what a trial is allowed to prove.

What to watch

The free FDA calendar at pdufa.bio (data through September 8) puts several near decisions on the board, including Ultragenyx’s UX111 (ABO-102) PDUFA on September 19 and a cluster of late-September dates spanning Merck’s Winrevair label work, Incyte and Mirum’s zilurgisertib, Novo’s Mim8, and Scholar Rock’s apitegromab. Treat each date as a deadline with a document trail, not as a forecast.

For mechanism readers, the open questions are sharper than the tickers. Does any Lp(a)-lowering program survive Horizon’s outcome lesson once full tables appear at a medical meeting. Can Etcamah’s ctDNA-guided switch hold up under confirmatory scrutiny and ordinary clinic workflow. Will Harbor’s noisy functional endpoint rewrite how DM1 programs pick their primary meters.