Tarlatamab (DLL3 BiTE) added to durvalumab plus carboplatin and etoposide versus durvalumab plus carboplatin and etoposide alone in first-line extensive-stage small cell lung cancer. A test of whether adding a DLL3-directed T-cell engager produces a meaningful increment on top of an already-active ICI + chemo backbone in the first-line window.
DeLLphi-312 adds tarlatamab, a DLL3-targeted BiTE, to the IMpower133-style backbone of durvalumab plus carboplatin and etoposide in first-line extensive-stage SCLC. The question is whether the DLL3 engager covers a structurally new step in the immune cycle on top of the backbone, or whether it stacks additional pressure on a step the backbone already covers.
The framework's calibration table places the backbone (durvalumab + platinum + etoposide) at approximately 68% ORR in 1L ES-SCLC. Adding tarlatamab produces a predicted ORR of approximately 70% — a delta of roughly 2 percentage points. This is below the 5-point delta the framework requires to call a WIN, and the confidence is low given the narrow window.
The structural reason for the small delta is that the 1L ES-SCLC window is already close to the ceiling of what can be achieved by engaging the tumour at the effector encounter step. Tarlatamab is an active molecule in later lines, where the backbone has already failed, but the first-line window leaves little room for an additional effector-step mechanism to produce a large increment. The framework calls this EQUIVALENT within a ±5 point band.
Each row below is the call the framework makes from the trial design alone, before the primary readout. DeLLphi-312 is actively enrolling (ClinicalTrials.gov NCT07005128, last updated April 2 2026). The phase 1b DeLLphi-303 supporting cohort on the same regimen (ESMO 2025, Lancet Oncol) reported an ORR of 71% — consistent with the ~70% framework call and inside the ±5pp EQUIVALENT band, though phase 1b is not a phase 3 readout and the primary OS endpoint has not been reported.
| Metric | Framework call |
|---|---|
| ORR experimental (tarla + durva + chemo) | ~70% |
| ORR control (durva + chemo) | ~68% |
| Δ experimental vs control | ~2pp (within ±5 band) |
| Framework call | EQUIVALENT, low confidence |
| Falsifier | OS HR delta outside [-5, +5] |
The falsifier: if DeLLphi-312 reports an OS HR delta outside the ±5 point band — in either direction — the framework's EQUIVALENT call is refuted. A large WIN would indicate that DLL3 engagement covers a structurally different step than the backbone in the 1L window; a large FAIL would indicate interference. The trial is active and the primary readout is pending.
The structural reasoning behind the call — which immune-cycle steps the regimen covers, which prior trials in the same structural class had already read out, and why the framework places DeLLphi-312 in that class — is in the full report. The full report also includes the framework's per-metric reasoning for each row in the table above and the implications for sponsors running similar combinations in adjacent indications.