DeLLphi-305: first-line maintenance tarlatamab plus durvalumab lifts median OS in ES-SCLC from 12.7 to 21.1 months, and bispecific TCEs formally enter first line

Data as of 2026-10-25 (Taipei time). [Fact] statements are sourced; [Inference] statements are this article's judgment.

1. One-line conclusion

[Fact] At DeLLphi-305's interim analysis, first-line maintenance tarlatamab plus durvalumab versus durvalumab alone gave mOS 21.1 vs 12.7 months (HR 0.42; 95% CI 0.32–0.54; P<.001) (Targeted Oncology). [Inference] The DLL3 T-cell engager will move from second line to first-line maintenance, but the trial enrolled only patients without progression after chemo-immunotherapy induction, so the number who benefit in the real world will be smaller than the total ES-SCLC population.

2. What happened

[Fact] DeLLphi-305 (NCT06211036) is the open-label Phase 3 trial reported as ESMO 2026 LBA3, enrolling ES-SCLC patients without progression after first-line platinum plus etoposide plus durvalumab. 563 patients were randomized 1:1 to tarlatamab 10 mg every 2 weeks plus durvalumab 1500 mg every 4 weeks (n=282) or durvalumab alone (n=281); about 20% had brain metastases and about one third liver metastases. The primary endpoint was OS, with PFS as the key secondary endpoint. Data cutoff was 2026-07-07; the interim analysis was performed at 243 events (61% information fraction), with median follow-up of about 16 months (Targeted Oncology).

[Fact] mOS was 21.1 months (95% CI 20.8–NE) vs 12.7 months (11.2–14.8), HR 0.42 (0.32–0.54; P<.001), 12-month OS 79% vs 53%; brain-metastasis subgroup HR 0.49 (0.26–0.93), liver-metastasis subgroup HR 0.41 (0.28–0.61). mPFS was 7.4 vs 1.9 months, HR 0.34 (0.28–0.42; P<.001); ORR 26.4% vs 4.1%, mDoR 12.7 vs 5.6 months. Grade ≥3 TEAEs 46.8% vs 40.6%, treatment-related grade ≥3 20.9% vs 10.8%; tarlatamab discontinuation for AEs 5.7%; treatment-related deaths 0 vs 2 (0.7%). CRS 49.6% (1 grade 3 case), ICANS 1.1% (Targeted Oncology).

For sponsors: from today the control arm for first-line ES-SCLC maintenance may be tarlatamab plus durvalumab; trials being planned that still use durvalumab alone as control may be outdated by the time they read out.

3. Why it succeeded

[Inference] Timing of intervention is the main reason. Control-arm mPFS with durvalumab alone was only 1.9 months (Targeted Oncology), meaning residual disease after induction was barely controlled, leaving ample room for improvement; after induction tumor burden is lower and the immune system has not yet been exhausted by multiple lines of therapy, which in theory is the best point for a T-cell engager.

[Inference] Manageable safety was also a prerequisite. CRS occurred in nearly half, but with only 1 grade 3 case, ICANS 1.1%, treatment-related deaths 0 and discontinuation for AEs of 5.7% (Targeted Oncology); first-dose monitoring remains a barrier to wider adoption.

[Inference] Statistically, crossing the boundary at a 61% information fraction (Targeted Oncology) indicates a large effect size; open-label design has little effect on OS, but PFS could carry assessment bias, and reports do not say whether it was assessed by BICR. Randomization took place after induction (Targeted Oncology), so patients who progressed during induction are not in the trial, and the screen-failure rate could not be found.

4. Regulatory and commercial implications

[Inference] Meeting the OS primary endpoint at the interim analysis (HR 0.42, Targeted Oncology) is usually enough to support filing; whether full approval follows depends on FDA review. "A new first-line maintenance standard" was the presenter's conclusion (Targeted Oncology).

[Inference] Competition should be viewed by treatment setting. DeLLphi-305 occupies the post-induction "maintenance" setting; BioNTech's pumitamig is being tested in first-line SCLC "induction" (pumitamig plus chemotherapy) in ROSETTA Lung-01 (BioNTech); GSK's B7-H3 ADC ris-rez currently has data in "relapsed" SCLC, enrolling patients including those previously treated with tarlatamab (GSK). [Inference] A successful new induction drug would change the makeup of patients entering maintenance; drugs for relapse will need to show efficacy in tarlatamab-pretreated patients.

5. Recommendations for the next trial design and OncoLattice

[Inference] First, update the control arm. Any first-line ES-SCLC maintenance trial being planned should consider including tarlatamab plus durvalumab in the control, given its OS improvement of HR 0.42 (Targeted Oncology).

[Inference] Second, test outpatient CRS monitoring. CRS incidence was 49.6% but almost entirely low grade (Targeted Oncology); trials could compare outpatient observation plus remote monitoring after the first dose with inpatient observation, lowering cost and letting community sites take part.

[Inference] Third, cover patients who progress during induction. DeLLphi-305 enrolled only those without progression after induction (Targeted Oncology); these patients are a clear gap in trials.

[Inference] Fourth, site strategy. T-cell engagers require the ability to manage CRS and ICANS (tocilizumab access, 24-hour emergency care), which OncoLattice should make a hard requirement when matching sites.

6. Open questions and what to watch

The final OS analysis will read out in later analyses of NCT06211036 (Targeted Oncology); whether PFS was assessed by BICR, quality of life and the screen-failure rate will most likely appear in the LBA3 full paper, whose journal and date have not been announced; for now there are only ESMO 2026 abstract reports (Targeted Oncology) and a paywalled Endpoints report (Endpoints). Amgen's filing and FDA decision dates have not been announced.

The next thing to watch is the screen-failure rate: it determines what share of first-line ES-SCLC patients can actually use this regimen, and the enrollment potential of trials of new induction drugs.

Evidence limitations

Efficacy and safety figures come from Targeted Oncology's report (Targeted Oncology); the Endpoints article is paywalled and only its lead could be read, which is consistent with the former (Endpoints).

Advice for readers: sponsors planning first-line ES-SCLC trials should re-estimate control-arm assumptions and sample size now using tarlatamab plus durvalumab's mOS of 21.1 months (Targeted Oncology), without waiting for the full paper; but estimates of CRS monitoring resources should wait for the full paper's grading and timing distribution.

Note: differences from the original relay

The relayed figures match. Additions: of the 49.6% CRS, only 1 case was grade 3; the upper CI bound of the experimental arm's mOS is not estimable.