LITESPARK-012: adding belzutifan or quavonlimab to Keytruda plus Lenvima both failed, and triplet stacking in first-line ccRCC has reached its end

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

1. One-line conclusion

[Fact] Neither experimental arm of LITESPARK-012 beat Keytruda plus Lenvima: the belzutifan triplet had PFS HR 1.08 (95% CI 0.93–1.26; P=0.84) and OS HR 1.04 (0.87–1.25; P=0.66) (GU Oncology Now). [Inference] This is not a near miss; the numbers point the wrong way. Adding a third active drug to IO plus TKI in first-line ccRCC may have a net effect of zero or worse if it sacrifices the dose intensity of the backbone.

2. What happened

[Fact] Merck and Eisai announced on 2026-04-21 (US Eastern) that the interim analysis missed both primary endpoints (Merck); the ESMO abstract (3586O) made the full figures public around 10/21 (ApexOnco). This open-label Phase 3 trial (NCT04736706) randomized 1688 untreated patients with advanced ccRCC 1:1:1 to three arms: arm A, pembrolizumab 400 mg Q6W plus lenvatinib 20 mg plus belzutifan 120 mg (n=560); arm B, MK-1308A (pembrolizumab plus quavonlimab 25 mg) plus lenvatinib (n=562); and the control, pembrolizumab plus lenvatinib (n=566). IO was given for up to about 18 cycles (2 years); the dual primary endpoints were PFS by BICR and OS (Merck, GU Oncology Now).

[Fact] Data cutoff was 2026-02-04, with median follow-up of 43.6 months. Median PFS was 20.8 months in arm A, 20.7 in arm B and 23.7 in the control; HR for A vs control 1.08 (0.93–1.26; P=0.84), for B vs control 1.07 (0.92–1.25; P=0.82). Median OS was 52.0 months in arm A and not reached in arm B or the control; HR 1.04 for arm A (0.87–1.25; P=0.66) and 1.12 for arm B (0.94–1.35; P=0.90). ORR was 76.3% in the control and 69.0–69.1% in the experimental arms. GU Oncology Now reports all-cause adverse event rates of 51.4–51.6% in the experimental arms and 36.2% in the control (GU Oncology Now); [Inference] an all-cause AE rate of only 36% with IO plus TKI in advanced RCC is implausible, so this is likely a reporting error (for example, grade ≥3 or serious AEs), pending the original abstract. The authors note that more experimental-arm patients discontinued for intolerance, which may have limited drug exposure (GU Oncology Now). ApexOnco notes that control-arm mPFS was consistent with the registrational CLEAR trial (ApexOnco).

For sponsors: a control arm that performed exactly as in its registrational trial, and experimental arms whose ORR actually fell, show that a first-line RCC triplet must first prove it does not erode the backbone's dose intensity before it is worth taking to Phase 3.

3. Why it failed

[Inference] The most likely explanation is toxicity reducing dose intensity. The experimental arms had higher AE rates and more discontinuations for intolerance, and the authors themselves suggested this limited drug exposure (GU Oncology Now); lenvatinib alone already required dose reduction in 69% in CLEAR (Lenvima label summary attached to the Merck release), and adding belzutifan or CTLA-4 lowered the effective exposure to the backbone.

[Inference] The reversed ORR is a signal supporting this hypothesis. If the problem were simply insufficient additivity, ORR would be flat; experimental-arm ORR was about 7 percentage points lower than control (69.0–69.1% vs 76.3%) (GU Oncology Now), which fits "reduced exposure" better.

[Inference] Two other factors are overlapping mechanisms and a strong control. Downstream of HIF-2α, which belzutifan inhibits, is angiogenesis, and lenvatinib already blocks VEGFR potently, so additivity is limited; the control arm had ORR 76.3% and mPFS consistent with CLEAR (ApexOnco, GU Oncology Now), leaving little room for a third drug. The trial also did not use a biomarker to select a population likely to depend on HIF-2α (Merck).

4. Regulatory and commercial implications

[Fact] Belzutifan plus lenvatinib was approved by the FDA on 2026-09-24 for ccRCC after PD-(L)1, based on LITESPARK-011 (PFS HR 0.74; final OS analysis not significant, HR 0.85) (FDA). [Inference] Welireg's growth can only come from later lines, the adjuvant setting and VHL-related indications, with no near-term first-line prospects; the value of developing quavonlimab in RCC has dropped sharply.

[Fact] Arcus's first-line casdatifan program chose to compare against Inlyta plus PD-1 (PEAK-21) and Opdivo plus Yervoy (PEAK-20) rather than Keytruda plus Lenvima (ApexOnco). [Inference] This is a strategy of avoiding the strongest control, but LITESPARK-012's lesson on triplet toxicity applies equally. Eisai said the results reinforce Keytruda plus Lenvima's central place in first line (Merck); the SOC is unchanged.

5. Recommendations for the next trial and OncoLattice (GU oncology)

[Inference] First, triplet trials must include a dose-intensity analysis. More experimental-arm patients in LITESPARK-012 discontinued for intolerance (GU Oncology Now); new trials should make relative dose intensity a key secondary endpoint and randomize a reduced starting dose of lenvatinib already in Phase 2.

[Inference] Second, switch to sequential or adaptive designs, for example IO plus TKI induction followed by adding HIF-2α based on depth of response or ctDNA, instead of starting with three drugs; LITESPARK-011 shows belzutifan does have PFS benefit after IO (FDA), so the biological basis for sequencing exists.

[Inference] Third, select patients and sites. First-line HIF-2α trials should be restricted to molecular subgroups likely to benefit; kidney cancer trials also need sites able to manage lenvatinib hypertension and proteinuria and belzutifan anemia and hypoxia (belzutifan label warnings in Merck). When OncoLattice evaluates sites, it should look at TKI dose-adjustment experience and follow-up frequency, not just enrollment numbers.

[Inference] Fourth, control-arm choice. Arcus has shown how to avoid Keytruda plus Lenvima (ApexOnco); but avoiding the strongest control invites the criticism that it is not optimal therapy, so sponsors need to engage the FDA early.

6. Open questions and what to watch

Discontinuation rates by arm, relative dose intensity and AE definitions await confirmation in the full paper of ESMO 2026 abstract 3586O (GU Oncology Now); the journal and date have not been announced. The trial is registered as NCT04736706. IMDC risk subgroup data for the belzutifan arm could not be found; only the quavonlimab arm's intermediate/poor-risk subgroup PFS HR 1.12 and OS HR 1.17 were found (GU Oncology Now). Design details of casdatifan's PEAK-20 and PEAK-21 will come from Arcus announcements and ClinicalTrials.gov registrations (ApexOnco).

The next thing to watch is the discontinuation rate and dose intensity in the full paper: if failure is confirmed to come from reduced exposure, every RCC trial adding a third drug to IO plus TKI will need to re-evaluate its starting dose.

Evidence limitations

The figures come from GU Oncology Now's and ApexOnco's summaries of the ESMO abstract; GU Oncology Now's all-cause AE figure of 36.2% looks low and may be a reporting error (GU Oncology Now).

Advice for readers: sponsors citing LITESPARK-012's AE rates to estimate triplet toxicity should wait for the full paper's AE grading tables rather than using the reported 51% vs 36%; for control-arm assumptions, the CLEAR-consistent mPFS of 23.7 months can be used (ApexOnco).

Note: differences from the original relay

The relayed figures (PFS 20.8 vs 23.7, HR 1.08, OS HR 1.04) match. CIs and p-values added. The top-line failure was announced on 2026-04-21; what became public on 10/21 was the full ESMO abstract. Merck's press release itself contains no HR figures.