On Saturday in Madrid, Nilofer Azad of Johns Hopkins presented the trial Compass Therapeutics has bet its lead drug on in an oral session at the European Society for Medical Oncology congress. About a month earlier, the FDA had told the company the trial wasn’t enough. The agency recommended a new study showing a survival benefit before Compass files for approval, and Compass said it would file anyway. The results were published in the Journal of Clinical Oncology the same day, and Compass’s press release led with tumor shrinkage. The coverage followed. MedPage Today reported that the tovecimig doublet “boosts response and PFS,” which is true and says nothing about the endpoint the agency had asked about.
Tovecimig is a bispecific antibody that blocks DLL4 and VEGF-A, two signals tumors use to build a blood supply. COMPANION-002 enrolled 168 people with advanced biliary tract cancer that had progressed after one round of chemotherapy and randomized them two to one: 111 to tovecimig plus the chemotherapy drug paclitaxel, 57 to paclitaxel alone. Patients at this stage have few options and not much time. By Compass’s count, about 26,500 Americans are diagnosed with biliary tract cancer each year. The second-line therapies commonly used, none of them FDA-labeled for the disease, generally shrink tumors in about 5 percent of patients or fewer.
On the scans, tovecimig won. Blinded central review found meaningful tumor shrinkage in 18.0 percent of patients on the combination and 5.3 percent on paclitaxel alone (p=0.0228), roughly 13 more responders for every 100 patients treated. According to the company, the median response lasted 4.9 months, median progression-free survival was 4.7 months against 2.6 (a hazard ratio of 0.44), and disease control was 62.2 percent against 36.8 percent.
8.9 months against 9.4
| Group | Value (months) |
|---|---|
| Tovecimig plus paclitaxel | 8.9 |
| Paclitaxel alone | 9.4 |
Overall survival is the endpoint that says whether any of that bought patients time, and it went the other way. Median survival was 8.9 months on the combination and 9.4 months on paclitaxel alone, with a hazard ratio of 1.05 and a p-value of .78. The point estimate leans slightly toward the control arm, and the trial can’t distinguish it from no difference at all. More tumors shrank. The trial did not show that anyone lived longer because of it.
Compass blames crossover, and the argument has some force. Control patients could switch to tovecimig once central review confirmed their cancer had progressed, and 31 of the 57 did. By the company’s tally, 142 of the 168 patients, 85 percent, eventually received the drug. Crossover on that scale can wash out a survival difference.
The evidence Compass offers for that explanation is weaker than the explanation itself. Its most striking figure compares control patients who switched, with a median survival of 12.8 months, to those who didn’t, at 6.1 months. The company’s own release calls that comparison post hoc and non-randomized, and the flaw is built in. To switch, a patient had to live to confirmed progression and still be well enough for more treatment. Splitting the control arm along that line separates the sturdier patients from the frailer ones, and then the drug gets credit for the gap.
The more careful correction is a two-stage estimation model designed to adjust for crossover. It produced a hazard ratio of 0.85, with a 95 percent confidence interval of 0.59 to 1.26 and a p-value of 0.53. The estimate tilts toward the drug, but the interval runs from a meaningful benefit to a meaningful harm, and this analysis was also done after the fact. Compass wrote crossover into its own protocol, and crossover is now being asked to explain away the entire survival result.
| Result | Value (HR) | Lower bound | Upper bound |
|---|---|---|---|
| Two-stage estimation model | 0.85 | 0.59 | 1.26 |
What 13 extra responses cost
The safety table got far less attention. In the company’s data, grade 3 or worse treatment-emergent adverse events occurred in 93.5 percent of the combination arm and 64.2 percent of the paclitaxel arm. Adverse events led 27.8 percent of combination patients to stop treatment, against 11.3 percent of controls. Treatment-emergent adverse events leading to death occurred in 5.6 percent against 1.9 percent. Hypertension, the signature effect of blocking VEGF, hit 69 percent of combination patients, and fatigue hit 67 percent. Compass reports no new safety signals. “Treatment-emergent” means an event happened during treatment, not that the drug was shown to cause it, and the release doesn’t give the size of its safety population. Read as conservatively as you like, the gap between the arms is still wide.
| Event | Combination (%) | Paclitaxel (%) |
|---|---|---|
| Grade 3 or worse adverse events | 93.5 | 64.2 |
| Stopped treatment for adverse events | 27.8 | 11.3 |
| Adverse events leading to death | 5.6 | 1.9 |
Weigh it the way a patient might. Out of 100 people treated, roughly 13 more get a tumor response. About 29 more have a grade 3 or worse adverse event, and about 4 more have an adverse event during treatment that ends in death. The randomized comparison shows no added median survival to set against that. Nobody yet knows whether the extra two months before progression felt better or worse to live through, because the patient-reported quality-of-life analyses are still ongoing and, the company says, will be reported separately.
“Not the response we expected”
On September 22, Compass disclosed the agency’s advice in a release filed with the SEC. The FDA “recommended that Compass conduct a trial demonstrating a survival benefit” before a biologics license application. That is a recommendation, not a refusal, and the company’s answer tried to have it both ways. Chief executive Thomas Schuetz said, “While this is not the response we expected, we respect the FDA’s feedback.” The same release says Compass “does not believe that a new trial is warranted prior to submitting a BLA.” Targeted Oncology reported that the company plans to go ahead with the submission and hasn’t set a timeline. The drug already has Fast Track and Orphan Drug designations.
The FDA is usually faulted for waving drugs through on tumor measurements, so this request was conservative and pointed. Its own August 2025 draft guidance on overall survival puts “an emphasis on the analysis of overall survival as a pre-specified safety endpoint,” which is a way of checking that gains on a scan aren’t being paid for somewhere else. A trial in which fatal treatment-emergent events run about three times higher in the treated arm, and the survival hazard ratio sits above 1, is close to the case that guidance was written for. When the agency finally holds out for the harder endpoint, the sponsor’s objection deserves the scrutiny the agency usually gets.
There is precedent for what survival evidence looks like in this disease. The UK’s ABC-06 trial randomized 162 patients with cholangiocarcinoma or gallbladder or ampullary cancer and, as the ASCO Post reported, improved median survival from 5.3 to 6.2 months with second-line FOLFOX. The gain was less than a month, and it still changed practice, because survival was what the trial measured.
Compass now has a positive primary endpoint, two FDA designations, a JCO paper and a podium in Madrid. Its chief executive respects the agency’s feedback. His company has told securities regulators that it doesn’t plan to run the trial the agency asked for before it files.
Sources
- Compass Therapeutics, COMPANION-002 results release (October 24, 2026)
- Compass Therapeutics, SEC Form 8-K exhibit on FDA feedback (September 22, 2026)
- Targeted Oncology, Tovecimig improves PFS but misses OS end point in biliary tract cancer (April 28, 2026)
- Targeted Oncology, on Compass’s plan to submit despite the FDA’s recommendation (September 23, 2026)
- MedPage Today, Tovecimig-Based Doublet Boosts Response and PFS in Advanced Biliary Tract Cancer (ESMO 2026)
- FDA draft guidance, Approaches to Assessment of Overall Survival in Oncology Clinical Trials (August 2025)
- ASCO Post, Second-line FOLFOX vs active symptom control in advanced biliary tract cancer (ABC-06)