Exelixis's STELLAR-304 Faces a Fourth Slip Before H2-2026 RCC Readout
The Phase 3 trial hit its enrollment target and completed its 2025 primary completion date, but topline guidance has moved four times since 2025 as the model assigns 91.4% for meeting the endpoint against a competitive readout window.
Executive Summary
- Exelixis has revised its STELLAR-304 topline guidance four times since mid-2025, moving from H2-2025 to a conditional H1-2026 to Q2-2026 and now to H2-2026, even though the trial's primary completion date of July 1, 2025 has not itself changed. That pattern separates operational readiness from communication timing and is the fact investors should track next.
- Enrollment closed above target at 317 patients, up from an original 291-patient goal, and the trial has been active but not recruiting since July 14, 2025. The mechanics of running the trial are not the open question; the timing of the public data release is.
- The dossier contains no primary endpoint result, effect size, or safety data for STELLAR-304, so the model's 91.4% endpoint-met probability describes the statistical bar's historical clearance rate, not an observed treatment effect [AXL-MODEL]. Investors reading this figure as a forecast of clinical benefit magnitude would be misreading it.
- Zanzalintinib's first NDA, in metastatic colorectal cancer, is under FDA review with a target action date of December 3, 2026, meaning STELLAR-304's H2-2026 readout window overlaps with an active regulatory decision on the same drug. A positive kidney cancer readout during this period would reinforce the drug's late-stage momentum across two tumor types.
- No trial in the dossier tests a MET-targeted agent head-to-head against STELLAR-304 in this same non-clear cell RCC population, but Bristol-Myers Squibb's completed CheckMate-9ER trial, pairing nivolumab with the related MET/VEGFR2 inhibitor cabozantinib, is the nearest mechanistic precedent. That precedent set a bar for TKI-immunotherapy combinations in kidney cancer that STELLAR-304 will be measured against.
The catalyst
Exelixis expects topline results from STELLAR-304, a Phase 3 trial testing zanzalintinib plus nivolumab against sunitinib in previously untreated advanced non-clear cell renal cell carcinoma, in the second half of 2026 Press ReleasePress ReleaseMay 5, 2026. Chief Executive Officer Michael Morrissey said the company's research organization is "on track to deliver multiple milestones this year across our zanzalintinib pivotal development program," naming STELLAR-304 alongside STELLAR-303 as expected readouts. The trial is registered as NCT05678673 and is designed to support registration, with two co-primary endpoints: progression-free survival by blinded independent radiology review and objective response rate, both under RECIST v1.1.
Probability of SuccessBased on the AppliedXL Probability of Success model. For more information about the methodology, read the research here.
Timeline drift
The stated H2-2026 window is the fourth guidance point Exelixis has issued for this trial. The company first flagged a H2-2025 readout in May 2025, then shifted to a conditional H1-2026 window tied to study event rates in July 2025, then to Q2-2026 in January 2026, and now to H2-2026 as of May 2026. None of these shifts changed the trial's registered primary completion date, which has stood at July 1, 2025 since the study's original filing. That gap, more than a year between primary completion and the current topline guidance, reflects the time-to-event nature of a progression-free survival endpoint, which requires enough events to accrue before the data can be locked and analyzed.
Operational footing
Enrollment finished above its original target, reaching 317 patients versus an initial 291-patient goal, an 8.9% increase recorded when the trial moved to active, not recruiting status on July 14, 2025. The registry-based protocol stability proxy rates the trial 'Stable,' with a single eligibility-criteria amendment as its only substantive change beyond the enrollment update. Those facts argue the trial itself has not been operationally troubled; the repeated guidance revisions concern communication of the readout date, not conduct of the study.
What the model read rests on
AppliedXL's 91.4% endpoint-met probability is carried mainly by phase-level and endpoint-type historical success rates, patient-per-arm powering, and the trial's registrational flag, according to the model's driver mix, while scientific-plausibility and mechanism-of-action-fit features pull in the opposite direction [AXL-MODEL]. That composition, weighted toward structural and historical base rates rather than mechanism-specific evidence, plus a confidence label of LOW, means the number reflects a statistical prior about Phase 3 time-to-event endpoints in oncology more than a judgment on this specific PFS/ORR combination clearing its bar.
Regulatory and competitive backdrop
Zanzalintinib's first regulatory filing, for previously treated metastatic colorectal cancer, was accepted by the FDA and carries a target action date of December 3, 2026, placing STELLAR-304's topline window inside an active review period for the same drug. In kidney cancer specifically, Merck's LITESPARK-034 trial, initiated in April 2026 under the companies' collaboration, pairs zanzalintinib with Merck's belzutifan (WELIREG) in the same disease, meaning a STELLAR-304 result will be read alongside that second pivotal program. No trial in the competitive set tests a MET-targeted agent head-to-head against zanzalintinib in this same non-clear cell RCC population; the nearest mechanistic and clinical precedent is Bristol-Myers Squibb's completed CheckMate-9ER trial, which combined nivolumab with cabozantinib, a related MET/VEGFR2 inhibitor, against sunitinib in previously untreated advanced RCC.
This analysis was produced using AI-assisted reporting systems, AppliedXL data, and official public records. These systems undergo editorial review, quality checks, and regular audits by human experts. Errors may still occur, as with any automated system. Always consult the linked primary sources. Read our AI Editorial Policy.