Compass Therapeutics still hopes for full FDA approval of its DLL4 x VEGF-targeting antibody tovecimig, but it now has a big problem: the molecule has missed a key survival endpoint in its pivotal Companion-002 trial in second-line biliary tract cancer.
On Monday the study was revealed to have failed on overall survival – something Compass blamed on a high rate of crossover from the control to the tovecimig arm – but did hit on progression-free survival. Both were key secondary endpoints; the study earlier succeeded on its primary outcome measure, overall response rate, but the project’s chances of approval hinge on survival results.
During a call to discuss the data, Compass’s chief executive, Thomas Schuetz, said the company still intended to seek a full go ahead for tovecimig, but seemed to be preparing investors for the possibility that Companion-002 was no longer sufficient to secure this.
The trial might be enough for accelerated approval, Compass suggested without giving concrete details of any plans, but this path would introduce the need for a new confirmatory trial. The company's stock traded down 60% on Monday morning.
Poor Companion?
The phase 2/3 Companion-002 compared tovecimig plus paclitaxel, versus paclitaxel alone, in patients who had previously progressed on gemcitabine and platinum-containing chemotherapy.
Last April it yielded a 17% ORR among patients receiving the combo, compared with 5% for paclitaxel, a result said to have hit statistical significance with a p value of 0.031.
PFS and OS data had once been expected in the second half of 2025, but were pushed back to the first quarter of 2026, then to this month. Compass blamed the delay on fewer deaths in the study than initially expected, leading to hopes that tovecimig might be improving survival; however, the actual picture turned out to be more complicated.
The PFS result seems fairly clear cut: the combo group had a median of 4.7 months, versus 2.6 months with paclitaxel alone, giving a hazard ratio of 0.44 and a p value of less than 0.0001.
However, OS was confounded by 54% of control patients crossing over to receive the combo. In the intent-to-treat analysis the median came out as 8.9 months for the experimental arm, versus 9.4 months for control (HR=1.05, p=0.78).
Companion-002 data in second-line biliary tract cancer
| Tovecimig + paclitaxel | Paclitaxel |
|---|
| ORR* | 17% | 5% |
| Stats | p=0.031 |
| mPFS | 4.7 mth | 2.6 mth |
| Stats | HR=0.44, p<0.0001 |
| mOS | 8.9 mth | 9.4 mth |
| Stats | HR=1.05, p=0.78 |
Compass also presented a rank-preserving structural failure time (RPSFT) analysis, designed to adjust for crossover, but said this result was “largely uninterpretable”. The numbers were essentially the same as the ITT analysis, because “when the ITT hazard ratio is above 1, RPSFT cannot work”.
Therefore, Compass fell back on a post-hoc analysis comparing overall survival among control patients who had crossed over, versus those who hadn’t. In the crossover group they saw a 12.8-month median OS, versus 6.1 months in the latter group. However, when questioned about the baseline characteristics of the crossover group, Schuetz only said this was something the company was analysing.
At the very least, the latest results call into question the ideal sequencing of tovecimig plus paclitaxel, and suggest that the MAb combo could be reserved for late-line use, following paclitaxel monotherapy.
As for safety, Compass didn’t give overall figures, or information on dose reductions or discontinuations. Hypertension seemed the most worrying side effect, with 44% of patients in the combo cohort experiencing a treatment-related grade 3 or higher event.
Compass now plans to complete its analysis of the study within around one month, and hopes to have a “meaningful conversation” with the FDA around mid-summer. If all goes well it could submit the BLA application before the end of the year, and gain approval in the second half of 2027.
Others companies have managed to get drugs approved despite studies being confounded by crossover, with Novartis’s Pluvicto being a high-profile example. It's notable that Novartis's analysis, adjusting for crossover, did show a nominally positive result despite its original hazard ratio being above 1. Compass said it was looking into other statistical methodologies to adjust for crossover.
Novartis has much deeper pockets. At the last count Compass had $209m in the bank, enough to fund operations into 2028. A confirmatory trial could put a big dent in this, if things don’t go as smoothly as the company hopes.
