Editor’s Briefing: September 14, 2026

In Short. World Conference on Lung Cancer data turned the weekend into a sorting exercise for ADCs, bispecifics, and niche TKIs: GSK’s Hansoh-partnered B7-H3 ADC posted an 18.5-month survival mark in Chinese relapsed small-cell disease while Roche/MediLink’s rival won on survival versus chemo but trailed that bar; Enhertu’s first-line HER2-mutant lung bid won progression-free survival yet showed a worse early death trend; Cullinan/Taiho and GSK’s Jideytro pressed exon-20 and ROS1 fronts; AstraZeneca’s newly accelerated oral SERD Etcamah missed Serena-4; Definium’s LSD pill went three-for-three in pivotal anxiety as FDA psychedelic hearings opened; and Cellectis scrapped allogeneic CAR-Ts for the in vivo bandwagon.

---

B7-H3 ADCs under a China-to-global microscope

At WCLC, Hansoh’s phase 3 of GSK-partnered ris-rez (B7-H3 ADC) in relapsed small-cell lung cancer reported median overall survival of 18.5 months versus 10.3 months on topotecan, a primary-endpoint win from a 461-patient China-only trial. That number sits above Amgen’s Imdelltra phase 3 mark (13.6 months) and above the roughly 12-month median Merck/Daiichi cited from phase 2 for ifinatamab deruxtecan (I-DXd), already filed. Cross-trial comparisons are unreliable, especially China versus global enrollment, and GSK’s own global ris-rez pivotal is not due until 2027; the near-term argument GSK floated is safety differentiation, including no grade 4/5 interstitial lung disease in Hansoh’s readout despite grade 3 ILD rates similar to rivals. (Fierce Biotech; FirstWord Pharma)

Roche-partnered MediLink answered with phase 3 tam-peli (YL201) in China: median OS 13.3 months versus 9.4 months on topotecan, a clear win that still trails Hansoh’s headline on a cross-trial basis. MediLink’s grade 3 ILD rate was 0.9% with no worse events, versus about 3.9% and 4.4% for the GSK-Hansoh and Merck-Daiichi programs in their reported datasets, and versus fatal ILD/pneumonitis cases in the Merck-Daiichi midphase experience. The teachable object is not a single winner: it is how sponsors will sell global relevance of China-generated OS bars while ILD and sequencing against Imdelltra decide share. (Fierce Biotech)

Enhertu’s PFS win, OS shadow

AstraZeneca and Daiichi Sankyo’s Destiny-Lung04 put Enhertu against Keytruda plus chemotherapy in first-line HER2-mutant NSCLC and hit the progression-free survival primary: a 37% cut in progression or death risk and median PFS 14.3 months (six months longer than control), with objective responses 70% versus 44.5%. Interim overall survival, not yet formally tested, ran the other way: median 29.3 versus 33.1 months, a preliminary 15% higher death risk on Enhertu, with curves separating after about 20 months. Investigators flagged uneven subsequent HER2-directed therapy (39% of treated progressors on the Enhertu arm versus 72% on control) and Enhertu-related grade 5 ILD/pneumonitis events; adjudicated drug-related ILD/pneumonitis hit 20.8% on Enhertu versus 2.3% on control. AZ says it will take the package to regulators anyway. PFS-without-OS packages have delayed or derailed other oncology filings in recent years; this one arrives with a known platform toxicity in the foreground. (Fierce Pharma)

Exon 20 and ROS1: oral options press the niche

Cullinan and Taiho’s Rezilient3 interim in untreated EGFR exon 20 insertion NSCLC showed zipalertinib plus platinum chemo at median PFS 14.5 months versus 8.5 months on chemo alone, with objective responses 65% versus 40.3%. Overall survival remains immature (hazard ratio 0.72). Grade 3+ adverse events were common on the combo (87.1% versus 54.4%), described as mostly manageable hematologic toxicity from chemo; EGFR-related grade 3+ events were called infrequent. A second-line FDA decision is already expected next February; Taiho plans to share the first-line package with authorities. Cross-trial, the 14.5-month PFS sits above the 11.4-month figure that supported J&J’s Rybrevant first-line clearance, while J&J separately reported Papillon final OS of 34.3 versus 27.9 months for chemo alone. (Fierce Biotech)

Separately, GSK used Arros-1 to push Jideytro (from the Nuvalent buy) toward a TKI-naïve ROS1 NSCLC filing this year after a July approval in the post-TKI setting. Among 94 efficacy-evaluable TKI-naïve patients, objective responses reached 94% (88/94) with a 15% complete response rate; 86% of responders remained in remission at 12 months, and all 10 evaluable patients with CNS metastases responded intracranially (70% intracranial CR). Dose reductions and discontinuations for treatment-related events across a broader 532-patient safety pool were 11% and 1%. ROS1 is only about 1–2% of NSCLC, but long treatment durations are the commercial thesis against a field that has underperformed on toxicity. (Fierce Pharma; FirstWord Pharma)

Ivonescimab’s survival receipt

Akeso and Summit’s Harmoni-2 finally delivered a statistically significant overall survival win for PD-1xVEGF bispecific ivonescimab over Keytruda monotherapy in first-line PD-L1-positive NSCLC: median OS 30.8 versus 22.6 months (about 27% improvement), with a sharper 42% OS advantage in the PD-L1-high (TPS ≥50%) subgroup where Keytruda alone remains global standard. Squamous histology showed about 35% OS improvement. The China-run head-to-head was always debated for Western practice mix; the OS receipt is what PD-1xVEGF buyers (Summit’s global program included) needed after years of PFS-first skepticism. (Fierce Pharma)

Etcamah’s first-line miss, a week after the letter

A week after FDA accelerated approval for a novel ESR1-mutation switch use during first-line aromatase inhibitor plus CDK4/6 therapy, AstraZeneca’s oral SERD Etcamah failed Serena-4: no statistically significant progression-free survival edge versus anastrozole when both were paired with Ibrance in first-line ER+/HER2− disease (numerical PFS improvement only). That joins Roche’s giredestrant miss in the same setting and tightens pressure on other first-line oral SERD bets such as Olema’s Opera-02. The new accelerated label is not automatically undone; FDA still requires a redesigned randomized trial (protocol targeted by April 2027) plus Serena-6 survival follow-up. Galbraith framed Serena-4 as reinforcing ESR1 testing on first-line therapy. The class’s cleaner long game remains adjuvant and ESR1-selected later lines, not beating aromatase inhibitors up front in unselected disease. (Fierce Pharma)

Three pivotal wins and a hearing room

Definium’s Panorama phase 3 in generalized anxiety put DT120 ODT (100 μg LSD) against placebo and a 50 μg psychoactive comparator meant to blunt unblinding worries: HAM-A fell 9.8 points at week 12 on 100 μg versus 4.7 on placebo (5.1-point placebo-adjusted gap; 245 patients), with key secondaries hit and no new safety signals flagged. That follows Voyage in anxiety and a June major-depression pivotal: three phase 3 wins in roughly twelve weeks, with an FDA meeting planned in the fourth quarter toward an NDA. Dosing is a single monitored eight-hour session. The same Monday, FDA held a public hearing on psychedelic drug development amid analyst notes of broad stakeholder support and Compass Pathways’ COMP360 still the near-term approval watch via priority-voucher chatter. (Fierce Biotech; BioSpace; FirstWord Pharma; STAT News STAT+ Definium headline only)

Allo CAR-T yields to in vivo

Cellectis discontinued clinical-stage allogeneic CAR-Ts lasme-cel (B-ALL) and eti-cel (NHL), citing falling relapse pools after better frontline regimens, slower enrollment, and competition from bispecifics and in vivo CAR-T. The company had called 2026 a momentum year for allo products as recently as February. The pivot matches a wider industry migration toward editing CAR-T cells inside the patient rather than manufacturing donor banks. (Fierce Biotech)

Calendar friction and a radiopharma rumor

FDA extended Exelixis’s zanzalintinib-plus-Tecentriq colorectal review by three months, to March 2027, while requesting updated safety and efficacy data after a June miss on one overall-survival primary in a metastatic CRC subpopulation. William Blair still called rejection risk low on the broader phase 3 package. Separately, the Financial Times reported GE HealthCare is nearing a possible purchase of radiopharmaceutical developer Sofie Biosciences for as much as $1 billion; that remains an unconfirmed deal report. CDMO Argonaut sold its life-sciences/diagnostics arm (now Aluris Sciences) for an undisclosed sum to fund Carlsbad fill-finish expansion. (Fierce Biotech; Fierce Biotech; Fierce Pharma)

What to watch

pdufa.bio and today’s regulatory packet put Ultragenyx UX111 (ABO-102, Sanfilippo A) on September 19, Merck Winrevair (HYPERION) label work on September 21, Mirum/Incyte zilurgisertib (FOP) on September 26, Biofrontera Ameluz for basal cell carcinoma on September 28, and a September 30 cluster with Novo Mim8 (denecimig) and Scholar Rock apitegromab still listed as Upcoming on the free calendar. FirstWord and STAT+ headlines over the weekend claimed an earlier Scholar Rock SMA approval; free bodies were not opened here, so treat that as unverified against the still-posted PDUFA line. Roche Tecentriq adjuvant MSI-H colon follows October 9; Merck/Daiichi I-DXd in extensive-stage SCLC is October 10; Roche Enspryng in thyroid eye disease is October 15. Telix Pixclara’s September 11 date remains in Awaiting status on the packet.

Open mechanism questions from this tape: whether China-only B7-H3 OS bars (ris-rez, tam-peli) replicate in GSK’s and others’ global studies before I-DXd’s October decision; whether Destiny-Lung04’s OS imbalance is subsequent-therapy artifact or a first-line Enhertu liability the FDA will not ignore; whether zipalertinib’s exon-20 PFS can coexist with Rybrevant’s OS lead; whether oral SERDs abandon unselected first-line altogether after Serena-4; and whether Definium’s three pivotal wins plus a friendly hearing room translate into a clean anxiety NDA path without psychotherapy-label complexity.