Pyxis Called Its MICVO Data Positive. The Stock Fell Anyway. Read Slowly.

On the morning of September 9, Pyxis Oncology put out a press release calling its updated data positive, and within hours the shares fell about 7% — then another roughly 13% the following session. That gap between a headline and a tape says more about this stock than either one does alone. A 194%-and-counting year had already priced something into Pyxis, and the update did not deliver it.

The update belongs to micvotabart pelidotin, or MICVO, Pyxis's only significant asset: a first-in-concept antibody–drug conjugate that targets the scaffolding around tumors (the protein extradomain-B of fibronectin) rather than a protein on the cancer cells themselves. The setting is late-stage head and neck cancer — patients with recurrent or metastatic disease who have already failed a platinum drug and a checkpoint inhibitor, which is most of what works. On the face of it, the numbers in that group were good: a 36% confirmed response rate, 94% disease control, median progression-free survival of 6.2 months, and a 79% chance of being alive at 12 months in a population with essentially no good options left.

That is a genuinely encouraging readout. It is also not the readout the market had been trading on.

The number that shrank

Four and a half months earlier, in April, Pyxis had shown preliminary data at a 46% confirmed response rate. The stock's climb — 194% year to date, and the reason this data day even mattered — was built in part on that 46%. The updated cohort is larger: 33 efficacy-evaluable patients against the earlier 13. Larger is more honest. And when the denominator grew from 13 to 33, the confirmed response rate settled from 46% to 36%.

This is the old lesson of small single-arm oncology cohorts, and it is worth stating coldly. Early response rates on a handful of patients routinely overshoot, because the handful is small and the patients are chosen. Ten responses out of 13 patients looks like 77%; it is not. Here the shift was not a collapse — 12 of 33 responses is real activity in a hard-to-treat group — but the market had been paying for the spicier version of the truth, and the updated data replaced it with the slower one.

The tolerance bill

The second recalibration is tolerability. MICVO works by carrying a toxic auristatin payload into the tumor, and that toxicity shows up systemically. More than half of the 35 safety-evaluable patients had a grade 3 or higher treatment-related adverse event; 40% needed a dose reduction; roughly 14% stopped treatment because of side effects. Peripheral neuropathy — a disabling, lingering nerve toxicity common to this payload class — hit grade 3 in about 17% of patients, while ocular treatment-related adverse events occurred in about 29% of the cohort by grade 1/2.

Pyxis has already been adapting. In December 2025 it introduced capping of the dose for high-weight patients to blunt these events. That is the tell: the company is actively trading away some drug exposure to keep patients on it. None of this makes the drug a failure — the response data is real — but tolerability is the difference between an academic response rate and an approved drug that doctors actually prescribe. In a cancer where the comparator is a single old antibody or a taxane, a drug that a large share of patients cannot stay on carries a narrower commercial ceiling than the survival curve implies.

The road is long, and the money is building-funded

Everything now rests on confirmation that does not exist yet. Pyxis is planning a randomized Phase 3 ("Headliner") of roughly 500 patients in this same second-line-and-beyond setting, against investigator's choice of cetuximab, docetaxel, or methotrexate, with both response rate and overall survival as co-primary endpoints. It has not started. Enrollment is expected to begin mid-2027, after an FDA end-of-phase-2 meeting in the first quarter of 2027. In between there is one near catalyst — updated combination data with Merck's Keytruda in the fourth quarter of this year — but the single most informative forthcoming events (longer survival follow-up) arrive next year.

Meanwhile the company is financing the wait, and dilution is part of the cost. Pyxis held roughly $33 million in cash and short-term investments after a July private placement worth about $50 million up front (up to roughly $114 million if warrants are exercised), with burn running around $25 million a quarter. That financing bought a runway only into the second quarter of 2027 — not through a Phase 3 trial that has not yet begun. Some of the enthusiasm on September 5, when the stock rose more than 9% on the financing news, was relief about funding; a Phase 3 at this burn and this valuation will likely demand another raise. The $282 million market cap is buying an early-stage signal, unconfirmed tolerability priced into the dose, and a still-invisible confirmatory trial.

What the selloff is really saying

Put the pieces together and the market's reaction stops looking contradictory. The headline said "positive," and the data was positive on its own terms. But the update shrank the number the run-up was priced on, reminded investors the drug is hard to tolerate, and left the value of the whole thesis resting on a randomized trial that has not opened. None of that is fatal. In a patient group with so few options, a 36% confirmed response rate and 79% twelve-month survival is a legitimate Phase 3 bet.

The useful question for a watcher is not whether the drug works — early evidence says it plausibly does — but whether the current price was already asking for the drug to work essentially flawlessly. The drop since the data landed is the market revising that assumption toward a slower, more ordinary path: a smaller confirmed response rate, real side effects, a long gap to a definitive readout, and more capital to raise along the way. Until the Phase 3 starts and the survival curve matures, the 194% year remains the part of the story that is already spent, not the part that is still being earned.