There exist drug classes that seem, in retrospect, cursed. As these chemicals worm their way through the clinical trial system, they consume billions of dollars along the way, and squelch through thousands of sick patients. When finally it dawns on everyone how useless the whole endeavour was, the drugs life is at last cut short, nothing useful left in its destructive wake. The prototype here are amyloid-beta drugs. These are Alzheimer’s treatments that are widely perceived as immense disappointments, with the negative sentiment even leaking to the broader public. To be fair to these chemicals, the story here is a bit more complicated than the tabloids let on. Lots of amyloid research was not fake, and the drugs may genuinely be useful for early-stage Alzheimer’s. But they remain, regardless, disappointments.

Beyond amyloid-beta, which has been steadily disappointing for awhile now, there is one other such category of drug whose particular dance has just recently wrapped up. It may very well someday gets its chance in the spotlight, but it will take time. Because it—just like every other chemical in this class—shares a searing, burning radioactivity. You should not touch them. You should not suggest touching them. In fact, no serious person should touch them for years to come, because to do so will be to receive the scorn of other serious people.

What I am talking about are, of course, TIGIT drugs.

TIGIT emerged in the wake of boundless enthusiasm from over a century of grueling cancer immunotherapy research. Much of this work went nowhere, but a small fragment of it helped produce the most valuable molecule in existence: Keytruda (pembrolizumab). This drug was so astonishingly, grossly successful that it would be barely an exaggeration to credit Keytruda with creating a Big Pharma. Since its approval in 2014, it has saved millions of years’ worth of patient lives, and will likely continue to save millions more.

So, if you worked in pharma R&D in the mid-2010’s, and you were on the hunt for the next big thing, “something like Keytruda” was the most attractive thing on the board. And TIGIT drugs were supposed to be that.

An explanation of what TIGIT actually is would require you to hold roughly seven concepts in your mind at the same time, the names of which—in characteristic immunology fashion—are not helpful in the slightest. What is important to understand is that TIGIT is a particular protein, and theorized to be another immune-system brake. The aforementioned century of immunology research had already proven that these brakes mattered: Keytruda worked by blocking a different brake and allowing immune cells to attack tumors again. TIGIT seemed to offer the same promise, since tumors appeared to exploit it to quiet nearby immune cells. But this story was set to be even more intriguing. Unlike Keytruda’s target, TIGIT sat at an especially busy intersection of immune regulation. Blocking it might not merely release one brake, but rather two brakes and one accelerator, tilting the local immune environment towards such an absurdly anticancer direction that it was unthinkable that it wouldn’t be clinically effective.

So, the theory went: block TIGIT, or create an ‘anti-TIGIT’ drug, and you’ve got something even better than Keytruda on your hands.

Dollar signs appeared in the eyes of nearly every pharmaceutical executive upon learning this. Roche was the first here, their group establishing the above scientific observations, publishing them in a 2014 paper. "The immunoreceptor TIGIT regulates antitumor and antiviral CD8+ T cell effector function”. The molecule that emerged from this work was something called tiragolumab; the first anti-TIGIT drug to exist.

Its initial clinical debut was at ASCO 2020, a major oncology conference. There, Roche discussed the results of a 135-patient phase 2 trial in metastatic non-small-cell lung cancer, randomizing patients one-to-one to the standard-of-care, plus either tiragolumab or placebo. The combination produced a response rate of 31% versus 16% in the placebo. Tiragolumab seemed to work. Yes, it wasn’t a cure for cancer, but neither was Keytruda, which still managed to rake in nearly ten billion dollars a year. The FDA granted tiragolumab a breakthrough designation in January 2021 on the basis of that study, and, within a year, Roche began to spin up phase 3 trials.

Blood was in the water for TIGIT, and though Roche was first to bite, others followed. Merck had vibostolimab. BMS had BMS-986207. BeiGene had ociperlimab, for which Novartis paid $300 million in early 2021 for co-development rights. Arcus had domvanalimab, for which Gilead in late 2020 paid $175 million up front plus a $200 million equity position in the company. iTeos, a little Belgian immuno-oncology outfit, had something called EOS-448, which GSK licensed in mid-2021 for $625 million upfront. Everyone wanted a bite and was willing to pay for it.

Typically, with drug classes that have as much buzz as TIGIT did, companies like to run multiple trials in parallel, each one focused on a different cancer or patient subpopulation. This is to avoid a situation where your drug works spectacularly, it gets approved for the cancer subtype you tested it on, and then you have to watch on as your competitors’ drugs flood the remaining subtypes with their copycat chemicals. And the theoretical evidence for TIGIT was so strong, so overwhelming, that when combined with the promising phase 2 results, pushed Roche to go all in. At one point, they were running twelve concurrent Phase 2 and Phase 3 trials, each focused on a slightly different patient population, altogether covering ~5,000 human lives. This effort, which was branded ‘SKYSCRAPER’, represented one of the largest parallel-indication programs in modern immuno-oncology, its total costs likely running into the multiple billions.

In May 2022, the first crack showed. Roche reported that its first major Phase 3 trial, in first-line small-cell lung cancer (SKYSCRAPER-02), had missed on progression-free survival, or PFS. But this was not the end of the world. Small-cell lung cancer is a rather miserable disease. Relatively little works here anyway. This subtype has swallowed a long procession of drugs that excelled in other settings, so this was not viewed so much as a failure as it was an admirable, Hail Mary attempt that was almost assuredly not going to work out anyway.

But a few weeks later came a bigger problem: Roche’s flagship lung-cancer trial (SKYSCRAPER-01), tested on an ostensibly curable type of lung cancer, also missed on PFS. To be clear: they did not miss it by a lot. Roche would spend the next two years insisting that the values were in the right direction, just not at statistical significance.

Either way, the company demurred, the PFS metric is not what matters most. They were not wrong. PFS means something quite specific: from the start of the trial, how long did it take for a patient’s cancer to either grow on imaging or kill them. It is a useful data point, but it is ultimately a fuzzy surrogate of the metric that people actually care about: overall survival, or OS. How long did this patient live? Unfortunately, this metric takes years to read out and is confounded by whatever subsequent lines of therapy the patient picks up after the trial, so PFS is an often relied-on proxy metric.

And Roche believed that OS would ultimately exonerate tiragolumab.

And in August 2023, Roche ‘accidentally’ leaked data suggesting that overall survival had indeed improved on the drug. The stock ticked up on what in retrospect was the last uncomplicated moment of optimism in the TIGIT race.

On November 26, 2024, Roche reported the final OS analysis. The flagship trial had missed. The survival trend had narrowed to the point of insignificance, and the trial that was supposed to anchor the entire program—the indication on which Breakthrough Therapy Designation had been granted, the signal on which ten other trials had been launched—could no longer be the anchor.

But the flagship’s collapse wasn’t even the nadir. The nadir, really, was Roche’s worse-than-nothing readout in July 2024 (SKYSCRAPER-06), in the interim between the flagship’s PFS miss and its OS miss. It did not merely fail to show superiority to the standard of care, but was actively worse. Patients on tiragolumab died faster than the control group.

Everything began to unwind from here on out for Roche. A planned follow-up was canceled before it had really begun, and another was deprioritized. Over the subsequent year, the GI indications collapsed one after another, a locally advanced esophageal-cancer trial failed, a head-and-neck study was abandoned, and the last major Roche hope, in first-line liver cancer, missed PFS with no trend toward OS. The only success, awkwardly, was a trial in esophageal squamous-cell carcinoma (SKYSCRAPER-08), which produced statistically significant survival results. But by the time the full paper appeared in early 2026, Roche had already removed tiragolumab from its pipeline.

The TIGIT game, for Roche, had ended.

But what of the other players? Could it be that tiragolumab was the problem, and not the TIGIT hypothesis? Perhaps a different molecule, one still targeting TIGIT, would have worked.

After Roche, Merck was the second biggest believer in TIGIT. Remember when I said Keytruda had almost single-handedly created a Big Pharma? Merck is that pharma. And with their patent over Keytruda set to expire in 2028, they were the ones most interested—and best positioned—to own its successor. In their exuberance, they decided to match Roche: twelve parallel trials of their own, each one running an anti-TIGIT drug called vibostolimab.

The same pattern repeated. In May 2023, Merck’s melanoma trial was halted because vibostolimab was causing such a high rate of immune-related adverse events that patients were discontinuing therapy faster than any efficacy signal could accumulate. In August 2024, a small-cell lung-cancer trial was halted for OS futility, with the combination arm running worse than the control on both efficacy and safety. In December 2024, two more lung-cancer studies were abandoned halfway through the trial. And by 2025, Merck announced the discontinuation of the entire vibostolimab program.

But there was one last hope. What if the biological story here wasn’t complete? What if the original Roche paper, a decade back at this point, had gotten something wrong?

Every anti-TIGIT drug was structured like an antibody, a protein shaped like a ‘Y’. Only the top two segments—known as the Fab region—are actually interacting with TIGIT, while the bottom region—known as the ‘Fc’ region—interacts with an entirely separate set of receptors on an entirely separate set of immune cells. Typically, the two work in tandem. The Fab region binds to TIGIT-expressing cells, and the Fc region grabs onto nearby immune cells, forcing them to kill whatever the Fab region has attached to; a phenomenon called antibody-dependent cellular cytotoxicity (ADCC). Importantly, TIGIT is expressed on tumor cells and immune-suppressing cells, so ADCC was a reasonable thing to aim for.

But this could backfire. TIGIT was also expressed on the cancer-fighting T-cells that the drug is meant to support. So yes, these drugs may kill your enemies, but they will also kill your army, and the empirical net effect of this is little impact on how long a cancer patient will live. But it doesn’t need to be this way. While naturally-created antibodies always perform ADCC to some varying degree, there’s a lot more room for creativity with antibodies created in a vat: you can simply break the Fc region by mutating it. The result is ‘Fc-silent’ antibodies, which should still bind to TIGIT-expressing cells, but not kill them. Whether this would work at all was, luckily, testable. While Roche and Merck had spent billions on their Fc-active molecules, Arcus and Gilead had, in parallel, been developing domvanalimab: an Fc-silent anti-TIGIT antibody.

For most of 2024 and 2025, domvanalimab was carrying the collective hope of the entire TIGIT field on its shoulders, as essentially the last well-powered phase 3 program still running with a mechanistically distinct molecule. Starting in early 2024, the drug entered the crucial test of the Fc-silent hypothesis: a phase 3 trial in upper-GI cancers (STAR-221).

It did not work. In December 2025, the trial was halted.

And what of everyone else? The smaller bets around the edges were erased with even less ceremony. Novartis had paid $300 million in December 2021 for an option on BeiGene's ociperlimab, and in July 2023, Novartis looked at the emerging TIGIT phase 2 data across the field and simply handed the rights back, forfeiting the option fee in what turned out to be one of the better decisions any business-development team made that year. BeiGene continued on alone, and in April 2025 its phase 3 trial was terminated for futility on an interim OS analysis, ending the program. GSK's bet was the most expensive and, in some sense, the most depressing. In June 2021 they had paid iTeos, a Belgian immuno-oncology shop, $625 million upfront plus up to $1.45 billion in milestones for belrestotug. Again, zero benefit. GSK and iTeos mutually terminated the program, the collaboration, and any further enrollment in the study, all in the same press release. Two weeks later, iTeos announced that it was winding down, and was later bought by an outfit known for acquiring down-on-their-luck biotechs in hopes of selling off their parts.

Despite it all, TIGIT has not yet technically died. As one article puts it: ‘AstraZeneca becomes TIGIT’s last man standing’. Their drug is called rilvegostomig and is currently in eleven Phase 3 trials. Unfortunately, the core thesis of the drug is contingent on Fc-silence meaning anything, so it is difficult to imagine history pans out differently here.

In 2026, a BMJ Oncology analysis would give a clinical name to what had happened: “herding.” The authors estimated that nearly 49,000 patients had been enrolled in anti-TIGIT trials by pharmaceutical companies, at a cost of more than $3 billion, all because their fellow pharmaceutical companies were doing the same thing. The Fc-silent hypothesis had been tested and had failed. The Fc-active hypothesis had been tested and had failed. Combinations with every conceivable drug, across every conceivable demographic, across every conceivable cancer diagnosis—all of it, tested, and all of it, in the aggregate, failed.

Today, amongst many oncology investors and researchers, TIGIT has become close to a dirty word. Never, ever suggest touching TIGIT. It will not work.

After all this, one cannot help but ask: what had gone so wrong?

Unfortunately, the field does not yet have a clear answer. And it is unlikely there is one answer. As is often the case in biology, a target that sits at the busy intersection of many valuable things is that the very thing that makes it attractive as a target also makes it almost impossible to reason about cleanly. Perhaps TIGIT alone does not move the immune system in one direction, but instead tugs on a dense, locally contingent web of signals whose meaning changed from tumor to tumor, patient to patient. Perhaps modulating TIGIT is genuinely important, but would require the modulation of a half-dozen other targets for it to have the benefit that everyone expected of it. Perhaps TIGIT was actually transformative, but only for a very specific cohort of patient that the clinical trial apparatus is simply not built to discover at scale. Perhaps something else entirely.

What does feel likely is that TIGIT was not nonsense. The billions wasted was not an outgrowth of the publish-or-perish industrial complex, or something of the like. Genuinely intelligent theory was here, backed by years of genuinely intelligent wet-lab effort, and its eventual failure was, as far as I can tell, predicted by absolutely nobody. In fact, TIGIT was the golden child of what translational biology ought to look like. It had human genetics-adjacent plausibility, clean immunology, druggable extracellular geometry, a commercial precedent, and early clinical signal. It simply did not work.

Lots of people boil down the problem of drug discovery to toxicology, or target selection, or trial scalability. All these matter, yes. But sometimes the people behind a drug can do everything right, and it will still fail. Keytruda taught the pharmaceutical industry that the immune system had brakes, and it earned a place in the annals of cancer biology history for that. TIGIT taught the more humiliating, expensive lesson: not every brake is attached to wheels.

No posts