Another twist in the TIGIT saga

If you thought the TIGIT saga couldn’t get any more bizarre, it just did. Overall survival data from Roche’s Skyscraper-01 trial, a closely guarded secret under the study’s statistical integrity rules, was last night leaked in full on what appears to be a Roche portal.

The data relate to Skyscraper-01’s second interim analysis, known to have been passed around February without triggering stopping criteria, and the key takeaway is that they show a much more positive survival benefit for Roche’s anti-TIGIT MAb tiragolumab than many had expected. However, this is a highly embarrassing blunder, and it must be hoped that it does not invalidate Skyscraper-01 entirely.

The study had already blown up last year, failing on progression-free survival at first interim analysis. But the focus remained on the key secondary endpoint of overall survival, which was said to be immature; importantly, nothing had been revealed about alpha allocation in Skyscraper-01, meaning that neither the p value bar that PFS missed – nor the one OS needs to clear – was known.

Key test

Skyscraper-01 is a key test of the TIGIT approach, being the most advanced trial of this mechanism. It tests tiragolumab plus Tecentriq versus Tecentriq in front-line NSCLC patients expressing PD-L1 at ≥50%, but given Roche’s body language, and the inability of Gilead/Arcus to show convincing TIGIT data, had been written off by many analysts.

This was reinforced by the fact that the February 2023 date for possible readout of Skyscraper-01’s second interim analysis came and went without an announcement from Roche, meaning that the trial’s independent data-monitoring board had analysed the result and decided that it did not warrant a halt for efficacy (or futility).

Last night’s leak changes the dynamic considerably. It appears that an internal powerpoint presentation from Roche’s steering committee meeting, detailing the full OS interim analysis from Skyscraper-01, was made available on the internet. The leak was highlighted in a note to clients by Evercore ISI’s Umer Raffat, and confirmed this morning by Roche as “an inadvertent disclosure”.

So what does this reveal? Most importantly, an impressive looking OS benefit for the TIGIT combo, which yielded a 6.2-month interim OS benefit versus Tecentriq alone, with a 19% reduction in risk of death across the trial. The OS data from the first interim analysis were also revealed, and suggest that the survival curves are widening with time; the hazard ratio improved from 0.82 to 0.81.

However, one caveat is that Tecentriq monotherapy is underperforming the Impower-110 trial, which showed mOS of 20.2 months in PD-L1 ≥50% expressers. And the Skyscraper-01 OS data are immature: the hazard ratio’s upper bound is above 1.00, the p value is 0.08, and a close look at the survival curves reveals plenty of censoring, much of it before the interim point and unevenly split.

Source: Evercore ISI.

Patients censored at a given time point can still be analysed at subsequent visits, meaning in practice that the medians in Skyscraper-01 could yet change. 

The leak also hinted at the alpha allocated to the second interim analysis, saying the p value bar to be cleared at this point was 0.0336. For comparison alpha of 0.049 had been allocated to OS in the failed Skyscraper-02 SCLC study. This is not to suggest that sequential analyses will be the same in both trials; it can still only be guessed how much alpha has already been spent in Skyscraper-01, and how much remains for final OS.

And could things get worse for Roche? The Skyscraper-01 leak represents a serious breach of data integrity, and one possibility is that it results in a formal statistical penalty being imposed, significantly raising the bar for final OS success. 

Alternatively it could bring the question of bias into the trial, again undermining a final readout. However, OS is a hard, objective endpoint, and whether bias has been introduced would also depend on how many patients are on trial at the point of the leak. The powerpoint states that around 25% were on study at the interim analysis, which related to a November 2022 cutoff.

That said, Roche traded up 4% today, while two other TIGIT players, Arcus and iTeos, each opened up about 30%. Make no mistake: TIGIT is in play again.

The evolution of Skyscraper-01