PDS Biotechnology's confirmatory head and neck cancer trial carries an actual completion date of August 25, 2026 and an actual enrollment of 12 patients. No results, no announcement of success — a registry field, and a number that small. Six weeks later the same company's Phase 2 data landed in JAMA Oncology with a median overall survival of 39.3 months.
A 39.3-month median is the kind of number oncology meetings stop for. In first-line HPV16-positive recurrent or metastatic head and neck squamous cell carcinoma, the pembrolizumab arm of KEYNOTE-048 reported 12.3 months. The pembrolizumab control arm of LEAP-010 reported 17.9.
Then the registry entry for VERSATILE-003, the randomized Phase 3 that was supposed to test whether 39.3 was real, reads in full: discontinued solely based on financial constraints, not related to the safety of participants, the investigational product, or the conduct of the study. Twelve patients. Planned accrual was never close to that.
VERSATILE-002 gave 53 checkpoint-inhibitor-naive patients a two-part vaccine called PDS0101 alongside pembrolizumab, and reported 33.96% objective response and 39.3-month median overall survival. Every patient got the combination. Nobody got pembrolizumab alone, so the trial cannot say how much of that survival came from Keytruda.
The randomized trial that could have answered this was closed in August 2026 with 12 patients enrolled, five weeks before the paper appeared.
In July we covered Merck's personalised mRNA cancer vaccine, where a randomized Phase 2b gave a 49% cut in recurrence risk for melanoma patients. Same partner, same checkpoint backbone, a far cleaner design. This is the harder version of the same question.
The trial that produced the number had no control arm
VERSATILE-002 ran from March 29, 2021 to May 15, 2025 across 26 oncology centers in the US, UK and Ireland. It was open-label and single-arm, using a Simon two-stage optimal design run separately for each cohort. Eighty-eight patients were enrolled, 87 were treated, and 75 formed the modified intention-to-treat population used for efficacy.
PDS0101 is a therapeutic cancer vaccine on the company's Versamune platform, built from a proprietary R-DOTAP lipid adjuvant paired with an E6/E7 peptide pool drawn from the HPV16 oncoproteins. It was given subcutaneously in cycles 1 through 4, and again at cycle 12, always layered on top of the checkpoint backbone.
The statistical bet was explicit. In the checkpoint-inhibitor-naive cohort the trial tested a null response rate of 17% against an alternative of 33%. In the resistant cohort it tested a null of 5% against 20%. That is a real pre-commitment to a threshold, and the first cohort cleared it. It is also a one-sided design with 80% power against a bar the trial itself chose.
Read the response rate carefully, because the two review methods disagree. Central review, the prespecified endpoint, found 6 complete and 12 partial responses among 53 patients: 33.96%. Investigators found 5 complete and 14 partial responses, 35.85%. Disease control ran 60.4% by central review and 77.4% by investigator assessment. The paper leads with 33.96%. The company's September 22 release cites 36% by investigator review alongside 34% by central review, and the 77.4% disease control rate it markets company-wide is the investigator-assessed number, not the 60.4% on the paper's central review.
The gap is not misconduct. It is what happens when response assessment in a single-arm study with no adjudicated control is done twice by different people.
| Measure | ICI-naive (n=53) | ICI-resistant (n=22) | Pembrolizumab benchmarks |
|---|---|---|---|
| Objective response rate | ✔ 34.0% central review | ✗ 0% | ◐ 19–25% (KEYNOTE-048, LEAP-010) |
| Median progression-free survival | ✔ 5.3 months (95% CI 2.10–9.00) | ✗ 2.0 months (95% CI 1.4–2.1) | — |
| Median overall survival | ✔ 39.3 months (95% CI 23.90–NE) | ◐ 14.8 months (95% CI 8.5–26.0) | ✗ 12.3 / 17.9 months |
| Prespecified primary endpoint | ✔ met | ✗ not met | — |
Central-review efficacy in the mITT population versus historical pembrolizumab-based trials. Cross-trial comparison; the authors state it should be read cautiously. Source: JAMA Oncology, 2026.
Single-arm Phase 2 survival
95% CI 23.90 to not evaluable, so the upper end is unknown. 12-month OS 78.5%, 24-month OS 67.6% at median follow-up of 24.4 months. · JAMA Oncology, 2026
The cohort that got 0% answers the real question
Twenty-two patients came into VERSATILE-002 already failed by an immune checkpoint inhibitor. They received the same combination. None of the 22 responded.
Zero partial responses. Zero complete responses. The prespecified primary endpoint in that cohort was not met, and the paper says so plainly. The design had anticipated it: the trial was going to add 20 more resistant patients if it saw 2 or more responses in the first 21. It saw none.
PDS Biotech's own framing treats this as consistency rather than failure — checkpoint inhibitors work best early, so failure after prior exposure is expected. That reading is fair. It is also the whole investment case in one number. The value of PDS0101, if there is any, sits entirely in first-line use. It cannot rescue patients who have already failed the backbone drug it is combined with.
That narrows the addressable setting to first-line HPV16-positive disease, where pembrolizumab alone already works reasonably well and where PD-L1 expression confounds every comparison. The authors concede it directly: without a randomized design, the analysis cannot determine whether PD-L1 status modified PDS0101's contribution beyond pembrolizumab's known role. That single sentence is the honest core of the paper.
The safety profile is the least contested part. 13.8% of patients had grade 3 or higher treatment-related adverse events, no grade 5 events were recorded, and the single grade 4 event came a year after PDS0101 therapy. Injection-site reactions hit 65 of 87 patients, all grade 1 or 2. The company compares its 13.8% against a published 17% for Keytruda alone in the same CPS ≥1 population, which is a fair apples-to-apples framing of the same checkpoint backbone.
Confirmatory Phase 3, closed
Randomized 1:1 against pembrolizumab alone, start date May 30, 2025, status TERMINATED as of August 2026. · ClinicalTrials.gov NCT06790966, 2026
Twelve patients is the number to hold onto. A confirmatory Phase 3 started on May 30, 2025 and closed in August 2026 with an enrollment of 12. The registry was last updated on August 31 and posted on September 2, a fortnight before the Phase 2 paper appeared in print. Fifteen months of running, and the study whose whole purpose was statistical certainty never got past the first dozen participants.
Nothing in the registry suggests anything went wrong clinically. The why-stopped field names money and nothing else, excluding safety, product and conduct by name. For an issuer on Nasdaq, that means the equity story needed a capital source other than its own balance sheet.
✔ What the case for PDS0101 looks like
+ A pre-committed response threshold that the first cohort cleared rather than a post-hoc endpoint
+ Safety indistinguishable from the backbone drug it is meant to augment
+ A defined niche in first-line HPV16+ disease where the treatment effect, if real, would be commercially valuable
Confirmation criteria: a randomized readout showing the combination beats the backbone alone on overall survival
✗ What the case against it looks like
− Median PFS of 5.3 months sits well below the survival figure, which is the signature of a large non-responder fraction holding the tail
− No control arm, so the backbone drug's own contribution is unmeasurable and PD-L1 status confounds the comparison
− The trial enrolled one fewer evaluable participant than planned, modestly cutting statistical power
Refutation criteria: a controlled first-line study that lands inside the confidence interval's lower half
Merck helped write the paper it is being judged against
The funding statement on the VERSATILE-002 paper names two sponsors: PDS Biotechnology Corporation and Merck Sharp & Dohme. Merck provided the study drug, the funding, and support for study design, data collection and analysis.
The role-of-funder disclosure runs longer than most. The sponsors participated in the design and conduct of the study, the collection and management of data, its analysis and interpretation, and the preparation, review and approval of the manuscript. This is standard language in oncology trials, where a PD-1 partner supplies the backbone and sits in the room. It is also, precisely, the reason a single-arm result needs an external reader.
The investigator disclosures follow the same line. Kevin Harrington, Robert Worden and the principal investigator each reported research funding or personal fees from PDS Biotech during the study. Dr Price reported grants from both PDS Biotechnology and Merck. Independent assessment of the survival claim has to come from outside that circle, which is exactly the randomized trial that no longer exists.
Merck supplied the drug, funded the study, ran the analysis and reviewed the manuscript. That is normal for a combination trial. It stops being a neutral fact at the moment the control arm becomes unavailable.
What $22.55 million with a first-refusal actually buys
On September 8, 2026, six weeks after closing the Phase 3, PDS announced a private placement of up to $22.55 million led by Dr Patrick Soon-Shiong, founder of NantWorks, an affiliate of Nant Capital. The initial closing brought $11.3 million in gross proceeds on September 14.
The financing came with an attachment. NantWorks received a one-year exclusive right to negotiate an exclusive licence to PDS0101. Same announcement, two facts, and they point in opposite directions.
For an investor, a first-refusal is a real option with a one-year clock on it. It says someone with the resources to run a global Phase 3 has already looked at the asset and wants first sight of the terms. It also caps the auction: no other partner can negotiate during that year, and a holder who wants out can wait out the 12-month window instead of signing at today's price.
PDS is explicit about what comes next. The company says it intends to pursue strategic partnerships or externally funded opportunities for Phase 3 development, and it has moved PDS0301 in metastatic colorectal cancer to the front of its own priorities. That is an honest description of a small-cap biotech that just cancelled its own Phase 3 and is now looking for someone else to pay for it.
Which leaves the awkward arithmetic. A first-line head and neck cancer Phase 3 is a nine-figure undertaking with a comparator arm, and the population is narrow enough that the recruiting sites are identifiable. Soon-Shiong's own company is building the kind of oncology infrastructure that could run it. The unresolved question is whether the first-refusal is a price signal about the data, or a substitute for the control arm the data needs.
PDS published its survival figure without a control arm, terminated the trial that would have supplied one, and then raised money from a partner who gets first refusal on the asset. Each step is defensible on its own terms. Together they describe a company that believes in its number enough to publish it in JAMA Oncology, and knows what that number is worth without a comparator.