Two Rejections, One Approval, No New Patients
On August 6, 2026, the FDA granted accelerated approval to Tudriqev — vusolimogene oderparepvec-wtpg, known for most of its life as RP1 — in combination with nivolumab, for adults with unresectable advanced cutaneous melanoma whose disease progressed on a PD-1-blocking antibody.1 As sentences go, that one is unremarkable. The thirteen months preceding it are not.
The same agency rejected the same drug, on substantially the same dataset, twice. A complete response letter in July 2025.2 Another in April 2026.3 Between them, Replimune cut roughly 60% of its workforce, scaled back US manufacturing, and watched its share price bottom out at $1.70.4
Nine days before the approval, the FDA's own review team published a 50-page briefing document that concluded the application "does not include an adequate and well-controlled investigation that demonstrates substantial evidence of effectiveness."5 Two days after that, an advisory committee voted 10 to 3 that the data were evaluable and clinically meaningful.6 A week later, the drug was on the market at $450,000 a course.4
There are two easy versions of this story and both are wrong. In the first, a scrappy biotech finally beat a capricious regulator. In the second, a captured agency folded under pressure. What actually happened is stranger and more useful to understand: the FDA approved the drug and simultaneously marked its headline efficacy number down by nine percentage points. The press releases spent two years saying 33.6%. The label says 24.2%. Both numbers are real. Only one of them is the regulator's.
That gap is not a rounding artifact. It is the whole argument, rendered as arithmetic. Let me walk you through it.
Sources: FDA approval letter and package insert; FDA CTGTAC briefing document, July 2026.5,7
A Cold Sore, Weaponized
Strip away the nomenclature and Tudriqev is a herpes virus that has been surgically disarmed and handed a set of weapons.
Replimune's scientists screened 29 clinical cold-sore isolates of herpes simplex virus type 1 and picked the one that killed tumor cells most efficiently: strain RH018A.8 That choice already distinguishes it from talimogene laherparepvec — T-VEC, sold as Imlygic — which was built on a lab-adapted strain.
Then they deleted two genes. ICP34.5 is the neurovirulence factor; removing it means the virus can no longer replicate efficiently in healthy neurons, but retains the ability to replicate in tumor cells, whose antiviral defenses are typically broken. ICP47 normally blocks the machinery that loads tumor antigens onto MHC class I molecules for immune display. Deleting it makes infected cells more visible to T cells, not less.
Then they added two genes. Human GM-CSF, a cytokine that recruits and matures dendritic cells, the immune system's antigen-presenting professionals. And GALV-GP R−, a fusogenic glycoprotein borrowed from gibbon ape leukemia virus with its inhibitory R peptide clipped off. That last one is the genuine novelty. It causes infected tumor cells to fuse with their neighbors into giant multinucleated syncytia, which then die in a maximally inflammatory way — spilling tumor antigens into a microenvironment already flooded with GM-CSF.
The theory is that this converts a tumor into a personalized vaccine factory. Kill cells locally, in a manner that looks like infection rather than quiet apoptosis, and you may prime a systemic T-cell response that reaches lesions the needle never touched.
It is injected directly into tumors: 1 mL per centimetre of the largest lesion dimension, up to 10 mL per dose across all injected lesions, every two weeks for eight doses. The first dose is 106 plaque-forming units per mL; doses two through eight step up to 107. Nivolumab starts at week 3 and can continue about two years. Superficial lesions are injected by hand; deep and visceral lesions under imaging guidance.7
That last capability is the pitch. T-VEC's label carries an explicit limitation of use: it "has not been shown to improve overall survival or have an effect on visceral metastases." Tudriqev's whole thesis is that it does better than that.
The theory is elegant. Turn the tumor into a vaccine factory. Whether the vaccine actually travels is the question the FDA spent thirteen months refusing to concede.
Dr. Maren Cole33.6, 15.7, 24.2: Pick One
Here is the uncomfortable fact at the centre of this approval. The IGNYTE trial has produced at least six defensible response rates, and they range from 15.7% to 33.6% — in the same patients.
| ORR | Population | Analysis |
|---|---|---|
| 33.6% | n=140 | Modified RECIST, blinded independent review — the trial's primary endpoint9 |
| 32.9% | n=140 | Standard RECIST 1.1, blinded independent review9 |
| 32.7% | n=156 | Investigator-assessed, ASCO 2024 |
| 15.7% | n=140 | FDA's own primary re-analysis5 |
| 24.7% | n=89 | FDA's sensitivity analysis — 22 responders5 |
| 24.2% | n=91 | The approved label — 22 responders7 |
Notice the responder count in the last two rows. Twenty-two, both times. The label's denominator differs from the FDA reviewers' by two patients because of a technical definition — the agency counted patients with a non-injected target lesion, the label counts any non-injected lesion — but the numerator is identical. The approved label is, functionally, the FDA's sensitivity analysis. The 24.2% figure appears nowhere in the peer-reviewed publication.9
Why does restricting to patients with an uninjected lesion cut the number so sharply? Because of a problem the FDA stated with unusual bluntness:
Almost half of patients that the Applicant deemed to have had a response had all of their target lesions injected with RP1. Over half of patients with an Applicant-reported objective response did not have any non-injected target lesions to assess systemic antitumor activity.
FDA CTGTAC Briefing Document, July 2026Read that again slowly. In roughly half of the "responders," every tumor that shrank was a tumor that had a needle in it. That is not nothing — shrinking injected tumors is a real clinical effect, and for a patient with painful, ulcerating cutaneous disease it can matter enormously. But it does not demonstrate the systemic immune activation that justifies calling this an immunotherapy rather than a local ablative procedure. And for melanoma that has spread, local control is not what kills or saves you.
Replimune's counter-argument is not frivolous. Among responders, 79.0% of non-injected lesions shrank by at least 30%, and of 52 non-injected visceral lesions in responders, 65.4% shrank by that margin.9 Something appears to travel. The dispute is whether a lesion-level analysis restricted to responders can establish it, or whether that is simply describing the patients who were already doing well.
The durations diverge the same way. The journal paper reports a median duration of response of 33.7 months.9 A later data cutoff produced 24.8 months. The label says 14.1 months.7 Watch for a trap here: "14.1" also appears in the publication as the lower confidence bound of the 33.7-month median. Same digits, entirely different quantity. I expect to see them conflated in coverage for years.
What IGNYTE Actually Was
Design. Open-label, non-randomized, 51 sites across five countries. Patients had Stage IIIB–IV cutaneous melanoma with confirmed progression on at least eight consecutive weeks of anti-PD-1 therapy as their last prior line; 46.4% had also received anti-CTLA-4.9
Results: ORR 33.6% by modified RECIST and blinded independent review. Complete response 15.0%. Median PFS 3.6 months. Median OS 32.2 months, 3-year OS 45.5%.
Limitation: The registrational cohort was added by protocol amendment, not prospectively designed as pivotal. The independent review committee was implemented retrospectively. FDA had recommended a randomized design at a Type B meeting in March 2021; the company proceeded single-arm anyway.5
Two things about this trial deserve to be held in mind simultaneously, because commentary on both sides tends to hold only one.
The first: the responses are real, and unusually durable. Of the 22 responders in the label population, 86.1% of responses lasted at least six months and 54.6% at least twelve; the range extends past 34 months and some remain ongoing.7 Complete responses ran 15% in the full analysis set — and were strongly stage-dependent, 23.6% in Stage IIIB through IVM1a versus 5.9% in more advanced visceral disease.9 Durable complete responses in PD-1-refractory melanoma are not something the field can afford to shrug at.
The second: the median progression-free survival was 3.6 months. Roughly two-thirds of patients received eight rounds of intratumoral injection and got nothing. That number is in the publication, it is not disputed by anyone, and it is almost entirely absent from the promotional narrative.
Then there is the methodological problem, which is the part I find hardest to wave away. The protocol allowed re-injection of lesions after progression, and in a number of cases that re-injection occurred shortly before a scheduled response assessment. Some patients had lesions biopsied or surgically removed. Some responses were reclassified retrospectively on histology. The FDA's summary judgement:
The response methodology employed by the Applicant does not allow for a reliable estimate of the treatment effect of RP1 when combined with nivolumab, and it artificially increased the reported ORR and DOR.
FDA CTGTAC Briefing Document, July 2026On overall survival the agency was even shorter: the analysis "is not interpretable."5 Sundeep Agrawal of the Oncology Center of Excellence put the point to the committee directly, saying that any claim RP1 makes patients live longer cannot be proven without a concurrent control group and is not an evidence-based claim.6 The 3-year survival of 81.8% in responders versus 22.5% in non-responders is often cited as though it were a treatment effect. It is a prognostic observation. People whose tumors shrink live longer than people whose tumors do not, in every cancer, under every therapy, including placebo.
The One Time This Platform Faced a Control Group
RP1 has been tested against a randomized comparator exactly once, and it lost.
Design. The only randomized controlled trial ever run on the RP1 platform, in advanced cutaneous squamous cell carcinoma.
Results: Missed both co-primary endpoints. Complete response rate 38.1% versus 25.0%, at P=.040 against a prespecified threshold of P≤.025.
Limitation: Different tumor type, different checkpoint inhibitor partner, and a numerically favourable trend. But it is the only controlled evidence that exists, and it did not clear its own bar.
This matters more than it might seem, because the oncolytic virus field has a long and specific history of single-arm enthusiasm that evaporates on randomization. T-VEC's registrational trial, OPTiM, hit its durable response endpoint at 16.3% versus 2.1% for GM-CSF — and missed overall survival, 23.3 versus 18.9 months, P=.051.10 When T-VEC was combined with pembrolizumab in the 692-patient MASTERKEY-265 trial, it failed both co-primary endpoints and was stopped for futility.11
And in the population most relevant here — melanoma that has already progressed on anti-PD-1 — T-VEC plus pembrolizumab produced response rates of 0% and 6.7% in the two MASTERKEY-115 cohorts, with no complete responses.12
That is the honest frame for 24.2%. Set against the sponsor's own promotional 33.6%, the label number looks like a defeat. Set against the closest mechanistic comparator in the same clinical setting, it is roughly four to six times better. Both comparisons are fair. Only the second one tells you anything about the drug.
The wider field record is sobering. In 21 years, exactly four oncolytic viruses have held a standing approval anywhere in the world: H101 in China in 2005, T-VEC in 2015, Delytact in Japan in 2021 under a conditional and time-limited authorization, and now this. Pexa-Vec's Phase 3 in liver cancer produced a median survival numerically worse than sorafenib alone.13 Oncorus liquidated. Turnstone left the field. Targovax rebranded. One company's lead oncolytic asset was sold for four million dollars.
No New Data, Two New Directors
Between the second rejection in April 2026 and the approval in August, Replimune generated no new clinical trial. The June resubmission contained a three-year overall survival update — the analysis FDA had already called uninterpretable — plus responses to the complete response letter. IGNYTE-3 has not read out.3,5
Four things did change.
The number changed. The agency approved on its own restrictive analysis set rather than the sponsor's. That is verifiable by holding the briefing document table next to the label. It is also, in my reading, the most defensible thing the FDA did in this entire episode.
The leadership changed. Both complete response letters were issued while Vinay Prasad ran CBER and Marty Makary was Commissioner. Prasad resigned around May 1, 2026, three weeks after the second rejection; Makary followed the same month. The officials quoted in the approval announcement are acting appointees.1 I want to be careful with this observation, because correlation is doing a lot of work. Reporting by STAT indicated that the earlier resistance was driven substantially by Rick Pazdur, not Prasad, which cuts against the tidiest version of the narrative.
An advisory committee was convened — the first across three review cycles, scheduled three days before the goal date, after a lengthy agency-wide gap in advisory meetings.
The FDA said outside input mattered. Its press release notes the agency "considered input from clinical experts and patient advocates who emphasized the urgent need for effective options in this refractory population."1 That is an unusual thing to write into an approval announcement.
The committee itself was not enthusiastic so much as pragmatic. Lawrence Schwartz, a Memorial Sloan Kettering radiologist and a co-author of RECIST 1.1, voted yes while calling it "messy data."6 Hussein Tawbi of MD Anderson — who required both a financial interest acknowledgement and a participation waiver — argued the trial "was maybe not perfect, but it starts from the fact that the intent was not registrational," and that the therapy should be available until the Phase 3 reads out.6 Melinda Yushak of Emory voted yes and said plainly that "we do need the phase 3 data to fully answer this question."
The dissents were sharper. Mayo biostatistician Karla Ballman: "There is so much uncertainty to what that overall response rate is." And the patient representative, Diane Aronson: "Hope is important, but reality weighed into my vote."6
One more piece of context that should temper any story about a fully captured agency: the very same advisory committee, one day earlier, voted 9 to 3 against a different cell therapy.6 Whatever is happening at the FDA in 2026, indiscriminate permissiveness does not describe it.
$450,000, and What Else It Buys
Tudriqev lists at $450,000 per course before rebates.4 The only way to judge whether 24.2% justifies that is to look at what a patient in this exact position can otherwise be offered.
| Option | ORR | Burden |
|---|---|---|
| Lifileucel (TIL therapy) | 31.4% | $515k; boxed warning; tumor resection, lymphodepletion, high-dose IL-2, inpatient stay14 |
| Ipilimumab + nivolumab | 28% | Grade ≥3 treatment-related events in 57%15 |
| Tudriqev + nivolumab | 24.2% | Outpatient injection; grade 3/4 treatment-related events 12.9%; no treatment-related deaths7,9 |
| Lenvatinib + pembrolizumab | 21.4% | Oral; substantial VEGF-inhibitor toxicity |
| Nivolumab + relatlimab | 12.0% | Well tolerated |
| Ipilimumab alone | 9–13% | Grade ≥3 around 31% |
| T-VEC + pembrolizumab | 0–6.7% | The direct oncolytic comparator12 |
| Chemotherapy | 4–11% | Median survival under 9 months |
Read the table honestly and the case for Tudriqev is not the response rate. It is the ratio. Lifileucel produces a higher ORR, but requires surgery, lymphodepleting chemotherapy and high-dose interleukin-2, carries a boxed warning, and around half of referred patients never receive a dose. Ipilimumab plus nivolumab produces a higher ORR with grade 3 or worse toxicity in well over half of patients. Tudriqev is an outpatient injection with grade 3/4 treatment-related adverse events in 12.9% and, in 140 patients, zero treatment-related deaths.9
Nobody disputed the safety. The first complete response letter explicitly raised no safety concerns.2 The most common label adverse reactions are fatigue, fever, chills, musculoskeletal pain and nausea — the flu-like signature of an immune system being deliberately provoked.7
It is a live herpes virus
The label warns that providers, caregivers, close contacts, pregnant women and newborns must avoid direct contact with injected tumors, dressings and bodily fluids. Four herpetic events occurred in 140 patients. Antivirals require a 72-hour dosing delay.7
Injection is a procedure
Deep and visceral lesions need imaging guidance. Pneumothorax occurred in 3 of 52 lung injections, one requiring a chest tube. This is not a therapy that travels easily to community practice.7
The contribution problem is unsolved
Every patient also received nivolumab. FDA asked for a design that could isolate RP1's contribution in 2021 and never got one. How much of the 24.2% belongs to the virus remains genuinely unknown.5
Four years to an answer
IGNYTE-3, the randomized confirmatory trial of about 400 patients with overall survival as its primary endpoint, was roughly one-third enrolled at approval. The FDA letter sets completion at September 2030.7
On conflicts, the record is what you would expect and should still be said aloud. The trial was funded by Replimune. Five of the 29 authors on the pivotal publication are Replimune employees holding company equity, including Robert Coffin, who discloses that he is the inventor on all of Replimune's patents. Fourteen further authors disclose research funding, advisory roles or honoraria from the sponsor. The publication states that data were analysed by statisticians employed by the sponsor. Medical writing was sponsor-funded.9
Bishal Gyawali of Queen's University, writing before the approval in defence of the second rejection, made the sharpest version of the point — noting that many of the international oncologists who signed an open letter urging approval also held financial relationships with the company. His methodological objection stands regardless of what one thinks of the politics: with re-injection permitted just before response assessment, and surgical removal of some lesions, "the response rate, whatever is being quoted, is probably an inflated response rate."16
A Real Signal on Compromised Evidence
I keep returning to a single comparison. Tebentafusp won full approval in uveal melanoma on a 9% response rate, because a randomized trial showed people lived longer. Tudriqev received accelerated approval on 24.2%, because nobody yet knows whether people live longer. Response rate and survival benefit are different things, and the approval pathway is the FDA telling you exactly which one it has.
This is not Marketing Hype. The responses are biopsy-confirmed, unusually durable, and occurred in patients for whom chemotherapy delivers 4 to 11% and the closest oncolytic comparator delivered essentially nothing. The safety profile is the best in its category by a wide margin. Fifteen percent complete responses in PD-1-refractory melanoma is a real clinical finding and I am not going to pretend otherwise.
Nor is it Strong Evidence, and it is not close. There is no randomized trial. The contribution of the virus itself, as distinct from the nivolumab every patient also received, is undetermined — five years after the agency first asked for a design that could answer it. The survival data are, in the FDA's own word, uninterpretable. Half of the responders had no measurable uninjected disease at all.
Promising is the calibrated call: a genuine signal, in patients with genuinely few options, resting on evidence the reviewing agency itself judged methodologically compromised, with the real answer four years out. If you or someone you love is in this position, that combination may well be worth accepting. Just accept it with the correct number in hand. The label says 24.2%, not 33.6%, and the difference between those figures is not statistical noise. It is the difference between the sponsor's analysis and the regulator's.
Real, durable responses in a population with almost nothing else — but the FDA approved this on its own deflated number, not the company's, and the trial that could actually prove benefit does not finish until 2030.
- 1. US Food and Drug Administration. FDA grants accelerated approval to vusolimogene oderparepvec-wtpg in combination with nivolumab for melanoma. FDA Resources for Information on Approved Drugs. August 6, 2026. Accompanying press announcement quotes Karim Mikhail (Acting Director, CBER) and Megha Kaushal (Acting Deputy Director, OTP).
- 2. Replimune Group. Replimune Receives Complete Response Letter from FDA for RP1 Biologics License Application. Press release, July 22, 2025.
- 3. Replimune Group. Replimune Receives Complete Response Letter from FDA for RP1 Biologics License Application. Press release, April 10, 2026.
- 4. Replimune Group. Replimune Announces FDA Accelerated Approval of TUDRIQEV in Combination with Nivolumab. Press release, August 6, 2026. Pricing and workforce reduction detail per company disclosures and contemporaneous trade coverage (AJMC, STAT, BioSpace).
- 5. US Food and Drug Administration, Office of Therapeutic Products, CBER. Cellular, Tissue, and Gene Therapies Advisory Committee Briefing Document: BLA 125827, vusolimogene oderparepvec. July 2026, 50 pp. Contains FDA's independent ORR re-analysis (15.7%, n=140; 24.7%, n=89).
- 6. Cellular, Tissue, and Gene Therapies Advisory Committee meeting, July 30, 2026. Vote 10–3. Committee statements per FDA transcript and contemporaneous coverage (OncLive, BioSpace, CancerNetwork).
- 7. US Food and Drug Administration. TUDRIQEV (vusolimogene oderparepvec-wtpg) Prescribing Information and Approval Letter, STN BL 125827/0. August 6, 2026. Label ORR 24.2% (n=91); median DOR 14.1 months; postmarketing requirement completion September 2030.
- 8. Thomas S, Kuncheria L, Roulstone V, et al. Development of a new fusion-enhanced oncolytic immunotherapy platform based on herpes simplex virus type 1. Journal for ImmunoTherapy of Cancer. 2019;7:214.
- 9. Wong MK, Milhem MM, Sacco JJ, et al. RP1 Combined With Nivolumab in Advanced Anti–PD-1–Failed Melanoma (IGNYTE). Journal of Clinical Oncology. 2025;43(33):3589-3599. doi:10.1200/JCO-25-01346. Trial registration NCT03767348.
- 10. Andtbacka RHI, Kaufman HL, Collichio F, et al. Talimogene laherparepvec improves durable response rate in patients with advanced melanoma (OPTiM). Journal of Clinical Oncology. 2015;33(25):2780-2788.
- 11. Chesney JA, Ribas A, Long GV, et al. Randomized, double-blind, placebo-controlled, global phase III trial of talimogene laherparepvec combined with pembrolizumab for advanced melanoma (MASTERKEY-265/KEYNOTE-034). Journal of Clinical Oncology. 2023;41(3):528-540. N=692; both co-primary endpoints missed.
- 12. Chesney JA, et al. Talimogene laherparepvec plus pembrolizumab for advanced melanoma that progressed on prior anti–PD-1 therapy (MASTERKEY-115). European Journal of Cancer. 2024.
- 13. Gane E, Verset G, Meyer T, et al. Pexa-Vec plus sorafenib versus sorafenib in advanced hepatocellular carcinoma (PHOCUS). Liver Cancer. 2024;13(3):256-272. N=459.
- 14. FDA Approval Summary: Lifileucel. Clinical Cancer Research. 2025;31(19):4004. See also Chesney J, et al. C-144-01, Journal for ImmunoTherapy of Cancer 2022, and the five-year update, Journal of Clinical Oncology 2025, doi:10.1200/JCO-25-00765.
- 15. VanderWalde A, Bellasea SL, Kendra KL, et al. Ipilimumab with or without nivolumab in PD-1 or PD-L1 blockade refractory metastatic melanoma (SWOG S1616). Nature Medicine. 2023;29(9):2278-2285.
- 16. Gyawali B. Why the FDA Rejected RP-1 in Melanoma: A Data Deep Dive. Medscape. April 29, 2026.
- 17. IGNYTE-3 confirmatory trial. ClinicalTrials.gov identifier NCT06264180. Phase 3, randomized 1:1, approximately 400 patients, primary endpoint overall survival, versus physician's choice.
- 18. Nathan P, Hassel JC, Rutkowski P, et al. Overall survival benefit with tebentafusp in metastatic uveal melanoma. New England Journal of Medicine. 2021;385:1196-1206.