ALTAIR was supposed to be the trial that turned ctDNA from a prognostic marker into a prescription pad: find the patients whose blood shows molecular relapse after surgery, treat them with chemotherapy before the scan catches anything, and buy time. On Monday a Nature Medicine paper put the verdict in print. Read against the question the investigators wrote down in advance, the verdict was no.

The headlines say something else, because two months before the journal published, the public line had already moved off the failed primary analysis and onto a post-hoc reanalysis that did clear statistical significance. The story of ALTAIR is now two stories: what the trial showed, and what was selected to be shown.

The primary endpoint missed. Median disease-free survival was 9.30 months in patients given trifluridine/tipiracil and 5.55 months in patients given placebo, a hazard ratio of 0.79 with a 95-percent confidence interval that barely included no effect (0.60 to 1.05) and a p-value of 0.107. The trial enrolled 243 patients through the academic CIRCULATE-Japan platform between July 2020 and June 2023, all resected for stage 0 through IV colorectal cancer, all ctDNA-positive after curative-intent surgery and any standard-of-care therapy where indicated, all with no radiographic disease at the moment of randomization, randomly assigned to six months of an oral chemotherapy or six months of pills that looked like one. The investigators wrote down a primary endpoint, they reported the primary endpoint, and patients with no visible cancer were asked to swallow a toxic drug for a non-significant result.

The pivot came in January. A post-hoc blinded central radiographic review of the overall study population, presented at ASCO GI after a Natera investor release on January 6, slid the hazard ratio to 0.75, the confidence interval to 0.55 to 0.98, and the p-value to 0.041. That is the number you will see in company press materials and oncology-trade write-ups. It is the number that is not the primary endpoint.

The one place where the curve came alive was resected stage IV disease. The hazard ratio was 0.53, the p-value 0.012, and two-year disease-free survival ran 34 percent on trifluridine/tipiracil against 6.6 percent on placebo. That is a clean signal in a biologically distinct group, multiplicity-unadjusted and exploratory, and it deserves a confirmatory trial of its own rather than a press cycle pretending it was the answer ALTAIR asked.

The cost side is not subtle. Grade 3-or-higher adverse events landed in 73.0 percent of patients on trifluridine/tipiracil against 3.3 percent on placebo, with neutrophil-count decreases in 56.6 percent of the treated arm. Any-grade events ran 98.4 percent against 57.0 percent. Six months of pills, in a population that by definition has no imaging-detectable cancer at the moment of randomization, for a 3.75-month median DFS separation that, by definition, includes recurrence, second cancers, and death rather than relapse alone. Overall survival remains immature. That is the trade ALTAIR is offering, and it is worth scrutinizing whatever the p-value says.

The commercial geometry around the trial shapes how the result is being sold. Trifluridine/tipiracil is sold as Lonsurf by Taiho Pharmaceutical, which originated the drug in Japan and licensed rights to Servier in Europe and other markets outside the United States, Canada, Mexico, and Asia. Taiho supplied the drug and funding. The ctDNA assay that defined who was eligible is Signatera, made by Natera, the publicly listed diagnostics company whose January 6 investor release set the post-hoc number into circulation two days before the conference presentation and pitched ctDNA dynamics to Wall Street as validated. The trial itself was investigator-initiated through CIRCULATE-Japan, the academic platform, which is the cleanest piece of the picture. By the time the result reached a Monday-morning reader, the cleanest piece was no longer the visible one.

The honest sentence is the one in the abstract: trifluridine/tipiracil did not significantly prolong disease-free survival compared with placebo. The next honest sentence is that a subgroup of resected stage IV patients appears to benefit and deserves a confirmatory trial of its own. The third honest sentence is that detecting molecular relapse is now, technologically, easy, and the field still does not have a drug that reliably converts that earlier knowledge into longer survival. ALTAIR is the strongest randomized test of that proposition to date. It moved the needle. It did not move it as far as the press circuit, reading from the post-hoc tables, would like you to believe.

This should not change practice on the primary analysis alone. The journal publication and the favorable post-hoc language will travel further than the failed endpoint did.

Sources

  1. Nature Medicine – Post-adjuvant chemotherapy in ctDNA-positive patients with resected colorectal cancer: ALTAIR phase 3 trial (2026)
  2. ClinicalTrials.gov – ALTAIR trial registration (NCT04457297)
  3. Natera – Updated analyses from ALTAIR presented at ASCO GI (2026)
  4. Journal of Clinical Oncology – ALTAIR primary results, ASCO abstract LBA22 (2025)
  5. Nature Medicine – DYNAMIC-III ctDNA-guided adjuvant therapy in colon cancer (2025)