Roche's investigational TIGIT antibody tiragolumab loses head-to-head to pembrolizumab in a lung-cancer trial

Posted results for Roche’s Phase II/III SKYSCRAPER-06 trial mark one of the harder failures for the investigational anti-TIGIT class. In previously untreated advanced non-squamous non-small-cell lung cancer, the experimental antibody tiragolumab — added to atezolizumab (Tecentriq) and platinum chemotherapy — did not merely fail to beat the standard of care. It performed measurably worse than pembrolizumab (Keytruda) plus the same chemotherapy backbone, and more patients in the tiragolumab arm died. Tiragolumab is not an approved treatment; it is an experimental agent being tested in this trial.

The randomized, double-blind trial (NCT04619797) enrolled 542 patients, randomized 1:1 to tiragolumab plus atezolizumab plus pemetrexed (Alimta) and carboplatin or cisplatin (269 patients) versus placebo plus pembrolizumab plus the same chemotherapy (273 patients). Eligibility was restricted to non-squamous NSCLC and excluded tumors with EGFR mutations or ALK fusions, as well as ROS1 or BRAF V600E alterations, so the result speaks to a specific lung-cancer subtype, not lung cancer broadly. Both primary endpoints were analyzed in the full analysis set of all randomized patients, not a PD-L1-selected subgroup.

On the co-primary endpoint of investigator-assessed progression-free survival, median PFS was 8.31 months in the tiragolumab arm versus 9.89 months with pembrolizumab. The hazard ratio was 1.27 (95% CI 1.02–1.57), a confidence interval that excludes 1.0 in the direction favoring the control.

Overall survival and deaths both favored the control

Overall survival, the other primary endpoint, told the same story. Median OS was 18.89 months in the tiragolumab arm versus 23.10 months with pembrolizumab, for a hazard ratio of 1.33 (95% CI 1.02–1.73).

The safety record points the same way as the survival analysis. Across all causes, 54.3% of tiragolumab-treated patients died (146 of 269) versus 42.9% of pembrolizumab-treated patients (117 of 273). Serious adverse events affected 55.1% of tiragolumab-treated patients (147 of 267) versus 52.2% of pembrolizumab-treated patients (142 of 272). The higher mortality in the experimental arm is the most patient-relevant finding here: the tiragolumab combination was not only less effective on the survival endpoints but was associated with more deaths and more serious adverse events.

On objective response rate — a secondary, surrogate measure of tumor response (RECIST v1.1 complete or partial response) rather than survival or function — the tiragolumab combination produced responses in 50.2% of patients versus 56.6% with pembrolizumab. Because ORR captures only tumor shrinkage, not how long patients live or how they feel, it carries less weight than the two primary survival endpoints; here it happens to point the same way.

The 95% confidence intervals for both primary endpoints sit entirely above 1.0, and more patients in the experimental arm died — the tiragolumab combination did worse, not just no better.

For a head-to-head design, the choice of comparator matters: SKYSCRAPER-06 pitted the experimental regimen directly against pembrolizumab-based chemoimmunotherapy, the established first-line standard, rather than against chemotherapy alone. That makes the result a clean read on relative benefit, and the read is negative on both efficacy and safety.

The findings also align with broader setbacks across the TIGIT field, where adding the target to PD-(L)1 blockade has repeatedly failed to deliver the additive benefit that early-phase signals suggested. Association is not causation, and these are trial-level outcomes rather than guidance for any individual patient.