Background
Phase 3, randomized, double-blind, placebo-controlled trial. N=521 randomized. Previously untreated, locally advanced unresectable or metastatic non-small cell lung cancer with PD-L1-high tumours by central laboratory testing. Tests whether adding tiragolumab, an anti-TIGIT monoclonal antibody, to the PD-L1 inhibitor atezolizumab improves outcomes over atezolizumab alone. Built on an encouraging randomized phase 2 signal (CITYSCAPE) in the same PD-L1-high population.
Results
Interventions and follow up: Arm A: Tiragolumab 600 mg IV plus atezolizumab 1200 mg IV, 21-day cycles until progression, loss of clinical benefit or unacceptable toxicity (n=262)
Arm B: Placebo plus atezolizumab 1200 mg IV, 21-day cycles (n=259)
Primary endpoint: Co-primary — investigator-assessed PFS and OS in the PD-L1-high primary analysis set (22C3 assay)
mFollow up: 9.9 mo (IQR 6.2–13.8) at primary PFS analysis; 17.9 mo (IQR 6.7–39.0) at final OS analysis
Results: Median INV-PFS: 7.0 mo (95% CI 5.6–9.8) vs 5.6 mo (95% CI 4.4–7.0), HR 0.78 (95% CI 0.63–0.97; P=.02, did not meet the prespecified significance boundary)
Median OS: 23.1 mo (95% CI 17.7–28.8) vs 16.9 mo (95% CI 14.6–21.3), HR 0.87 (95% CI 0.70–1.10; P=.22)
ORR: NR in the primary report
Both co-primary endpoints: not met
Arm B: Placebo plus atezolizumab 1200 mg IV, 21-day cycles (n=259)
Primary endpoint: Co-primary — investigator-assessed PFS and OS in the PD-L1-high primary analysis set (22C3 assay)
mFollow up: 9.9 mo (IQR 6.2–13.8) at primary PFS analysis; 17.9 mo (IQR 6.7–39.0) at final OS analysis
Results: Median INV-PFS: 7.0 mo (95% CI 5.6–9.8) vs 5.6 mo (95% CI 4.4–7.0), HR 0.78 (95% CI 0.63–0.97; P=.02, did not meet the prespecified significance boundary)
Median OS: 23.1 mo (95% CI 17.7–28.8) vs 16.9 mo (95% CI 14.6–21.3), HR 0.87 (95% CI 0.70–1.10; P=.22)
ORR: NR in the primary report
Both co-primary endpoints: not met
Adverse events
Grade 3–4 (any): 41.2% (110 of 267) vs 33.8% (89 of 263)
Treatment-related deaths: 4 vs 2
Immune-mediated: per-event rates NR in the primary report
Discontinuation: NR
Treatment-related deaths: 4 vs 2
Immune-mediated: per-event rates NR in the primary report
Discontinuation: NR
Conclusions
Adding tiragolumab to atezolizumab did not produce a statistically significant improvement in either co-primary endpoint in previously untreated PD-L1-high advanced NSCLC. Both curves separated numerically — a 22% relative reduction in progression risk and a 13% reduction in death risk — but neither crossed its prespecified boundary, while grade 3–4 toxicity rose by roughly 7 percentage points and treatment-related deaths doubled. Anti-TIGIT blockade added toxicity without demonstrable benefit in this setting.
Key Limitations
A co-primary endpoint design splits alpha across PFS and OS, which is why an observed PFS hazard ratio of 0.78 with a nominal P=.02 is formally negative — a result that reads as favourable in isolation but fails the prespecified statistical plan, and a reminder to read the boundary rather than the P-value. Median follow-up at the primary PFS analysis was only 9.9 months. The comparator was atezolizumab monotherapy rather than pembrolizumab, the far more widely used single-agent PD-1 inhibitor in PD-L1-high NSCLC, and not chemoimmunotherapy — so the trial cannot say whether TIGIT blockade would add anything to contemporary standards. PD-L1-high selection shifted across the study, with the primary analysis set defined by the 22C3 assay after an earlier enrolment strategy, introducing assay-related population drift. No biomarker beyond PD-L1 was prespecified to identify a TIGIT-sensitive subgroup, so the trial cannot distinguish an inactive drug from an unselected population.
Clinical Context
Tiragolumab was the most advanced anti-TIGIT antibody in clinical development and SKYSCRAPER-01 was its lead registrational trial, carrying a hypothesis that attracted several billion dollars of industry investment across competing programs. The encouraging randomized phase 2 CITYSCAPE signal did not replicate. Roche disclosed the overall survival miss in November 2024; after concordant failures in SKYSCRAPER-06 in nonsquamous NSCLC and in hepatocellular, esophageal and head and neck trials, the company discontinued the tiragolumab program in 2025 and removed it from the pipeline. There is no regulatory filing and no guideline role for TIGIT inhibition in NSCLC or any other tumour. Single-agent pembrolizumab, or platinum-based chemoimmunotherapy, remains the NCCN, ASCO and ESMO first-line recommendation for PD-L1-high advanced NSCLC without targetable driver alterations. The value of this full publication is as the definitive negative record of a widely watched immunotherapy hypothesis and as a caution about advancing randomized phase 2 PFS signals into phase 3 without a validated predictive biomarker.