Tovecimig

COMPANION-002 randomized 168 participants 2:1 and met its primary endpoint of objective response rate, reported first as 17.1% (19 of 111) versus 5.3% (3 of 57) with p=0.031 and updated in the final analysis to 18.0% (20 of 111) with p=0.0228 after one participant initially classified as non-complete response and non-progressive disease was adjudicated a partial response by blinded independent central review. The trial was powered to detect a 23% absolute difference in response rate with 90% power and a two-sided alpha of 0.05, so the observed difference of about 12 percentage points is smaller than the effect the sample size was planned around, and the planned sample size of approximately 150 was exceeded by the 168 actually randomized. As of September 21, 2026 the complete dataset had not been published or presented; the sponsor stated that it was selected for an oral presentation at the European Society for Medical Oncology Congress in October 2026, so the results available are limited to sponsor announcements and a corporate presentation filed with the U.S. Securities and Exchange Commission, and are not peer reviewed.

Overall survival did not reach statistical significance. Median overall survival was 8.9 months with tovecimig plus paclitaxel versus 9.4 months in the control arm, hazard ratio 1.05 and p=0.78. The sponsor attributes this to crossover: 31 of 57 control participants (54%) crossed over after centrally confirmed progression, so 142 of 168 participants (85%) received tovecimig, and the control arm is not a tovecimig-naive comparator for survival. The prespecified rank-preserving structural failure time adjustment gave 8.9 versus 9.4 months with a hazard ratio of 1.13 and p=0.65, and the sponsor states that the assumptions of that method were not met and that its results here are largely uninterpretable. The crossover comparisons that show benefit, the 12.8 versus 6.1 month survival difference between crossover and non-crossover control participants and the pooled 9.8 month survival of all tovecimig-treated participants, are post hoc subset analyses of non-randomized groups; the same analysis showed crossover participants had progressed faster on paclitaxel alone, with median progression-free survival of 1.9 versus 3.6 months, indicating the two control subgroups differed.

The trial was conducted at 34 sites, all in the United States, and eligibility excluded participants who could receive an approved molecularly targeted therapy after first-line chemotherapy, so the results apply to the population without an actionable alteration rather than to all second-line biliary tract cancer. The registry record and the sponsor announcements describe different tovecimig schedules, day 1 and day 14 in the registry against days 1 and 15 in the press releases. Participant disposition, deaths, serious adverse events, discontinuations due to adverse events, disease control rate, duration of response, and the EORTC QLQ-C30 quality-of-life endpoint listed in the design publication had not been reported. Safety was reported for 108 and 53 participants rather than the 111 and 57 randomized, and grade 3 or higher hypertension occurred in 52% of the tovecimig arm against 6% of the control arm.

The phase 2 biliary tract cancer cohort was single-arm, open-label, and enrolled 24 participants at four sites, all in South Korea, so its 37.5% response rate has no concurrent control and was not estimated in a United States population. Responses were assessed by investigator review in the reported analysis rather than by the independent radiology review named as the registered primary outcome measure. Participants had received one or two prior systemic therapies, and the response rate differed markedly by line of therapy, 63.6% in the second line against 15.4% in the third line, so the estimate depends on the mix of lines enrolled. Two sources report different durations of response for the same cohort, a median of 9.4 months in the congress abstract and 6.9 months in the trial design publication, and median overall survival had not been reached at the August 12, 2022 data cut-off with a median follow-up of 11.8 months. Toxicity was frequent: treatment-related adverse events of grade 3 or higher were reported in 75% of participants, including one grade 5 pneumonia, and six participants (25.0%) discontinued because of a treatment-emergent adverse event. The study was terminated for a change of development plan, and the second stage of the Simon two-stage design was not conducted after the sponsor agreed with the U.S. Food and Drug Administration to move directly to a randomized trial, so the cohort was never expanded to the planned size. No baseline demographic table, quality-of-life outcome, or peer-reviewed full report of the cohort has been published.

COMPANION-003 studied tovecimig as monotherapy in metastatic colorectal cancer, not in biliary tract cancer, so it informs the safety profile and the DLL4 biomarker hypothesis rather than the indication. It was open-label and single-arm, enrolled 49 participants at sites in the United States only, and stopped after stage 1 because the protocol-defined criterion for progression to stage 2 was not met. Only 2 of 40 response-evaluable participants responded, and duration of response was estimated from those 2 participants, so the 14.4 week median is not a stable estimate. The sponsor reports response and disease control against a denominator of 40 response-evaluable participants while the registry posts outcomes against all 49 enrolled, so the two sets of figures are not interchangeable. The registry record specifies RECIST version 1.0 for the response definitions although the drug program otherwise uses RECIST version 1.1. No participant completed the study, 34 of 49 participants died, 7 withdrew, and 7 ended participation when the sponsor terminated the study, so follow-up was truncated. The comparison with anti-VEGF agents showing 1.0% to 1.5% response rates is a cross-trial comparison, and the statement that participants with DLL4-positive tumors did better is an unadjusted subgroup observation. Quality-of-life outcomes were not collected, and the results are available from the registry and sponsor filings rather than a peer-reviewed publication.

This is an investigator-sponsored phase 1/2 trial at a single institution with an estimated enrollment of 50 participants and no randomization or comparator arm. Its primary outcome measures are the incidence of adverse events and 6-month progression-free survival, so it is designed to establish tolerability and a maximum tolerated dose of CTX-009 added to gemcitabine, cisplatin, and durvalumab rather than to demonstrate efficacy. The population is first-line and previously untreated in the metastatic setting, which is a different population from the second-line population of COMPANION-002. Estimated primary completion is May 1, 2027, and no results had been posted or presented as of the September 9, 2026 registry update. The registry record does not state the doses or schedules of the four agents.

This is an investigator-sponsored, open-label, single-group phase 2 trial at one institution with an estimated enrollment of 25 participants, in metastatic colorectal cancer rather than biliary tract cancer. As of the June 23, 2026 registry update it was not yet recruiting, with an estimated start of August 31, 2026 and estimated primary completion of August 31, 2030, so no results exist. The single-group design without a comparator means the primary outcome of overall response rate cannot be attributed to the addition of tovecimig to FOLFIRI, and the planned quality-of-life measure is change in Eastern Cooperative Oncology Group performance status rather than a patient-reported instrument.

This is an investigator-sponsored phase 1b/2 trial at one institution in recurrent glioblastoma, not biliary tract cancer, with an estimated enrollment of 54 participants. The two arms enroll sequentially rather than concurrently, so the monotherapy and combination arms are not randomized against each other. The trial began on June 16, 2026 with estimated primary completion on June 30, 2029, and no results had been posted as of the June 22, 2026 registry update. The phase 2 primary outcome is the 12-month overall survival rate in each arm without a control group, so it will be interpreted against external benchmarks.

The first-in-human study was open-label, single-arm, and enrolled 45 participants at sites in South Korea only, so it supports dose selection rather than efficacy. No maximum tolerated dose was reached because no dose-limiting toxicities were observed across nine dose levels from 0.3 to 17.5 mg/kg, so the recommended phase 2 doses of 10 mg/kg and 12.5 mg/kg rest on activity and pharmacokinetics rather than on toxicity. Participants were heavily pretreated, with a median of four prior lines, and the tumor types enrolled were predominantly gastric (n=19) and colorectal (n=18) cancer; no biliary tract cancer subset result is reported, so the study provides no direct evidence for the indication. Two sources give different response-evaluable denominators for the same analysis, 39 in the congress abstract and 40 in the trial design publication, and the response rate of 18.8% at the recommended phase 2 doses is based on 3 responses among 16 participants. Only three confirmed partial responses occurred, all in participants whose archived tumor tissue showed high DLL4 expression, a finding from a small biomarker subset that was not prospectively tested. The study has been reported only as a congress abstract and, for pharmacokinetics, as a modeling paper using data from 31 dose escalation participants; no full clinical report has been published, and participant disposition and quality-of-life outcomes are not available.